A systematic review on the clinical effectiveness and safety of GAE for treating mild to moderate KO refractory to standard treatment or severe KO in non-surgical candidates was conducted according to the Cochrane Collaboration methodology,17 with reporting in accordance with the PRISMA statement.18 The protocol was registered in OSF [OSF osf.io/ytxvu]. Complementary, a systematic review of cost-effectiveness were performed following the Campbell-Cochrane Economic Methods.19

• 1)

  Information sources and search strategy

  The systematic review searches and subsequent identification of their included randomized controlled trials or non-randomized comparative studies was conducted using technological enablers from the Epistemonikos database, in April 2023. The results were automatically incorporated into the Epistemonikos L·OVE platform,20 where subsequent selection was conducted. An alert service for randomized controlled trials and non-randomized comparative studies was also created, which remained active until May 31, 2024.

  Additionally, to identify randomized controlled trials or other comparative studies not included in available systematic reviews, searches were conducted in MEDLINE (Ovid), Embase (Elsevier), CENTRAL (Wiley), and CINAHL (EBSCOhost), in October 2023.

  Economic evaluations were identified through searches in MEDLINE, Embase, and Web of Science, in June 2023. The terms from this strategy were combined with a strategy specifically designed by the University of York21 to identify economic evaluations.

  All search strategies were limited to studies published in English or Spanish within the last 10 years, as the earliest study on joint embolization for pain management in inflammatory musculoskeletal diseases was published in 2013.10

  Retrieved references were managed using Zotero 6.0.2322 and deduplicated using Deduklick,23 followed by manual removal of duplicates and study selection in Microsoft Excel.

  Furthermore, the bibliography of included articles was manually examined, and studies citing the selected studies were verified through Google Scholar.

  Full search strategies for all database searches are provided in Appendix A in Supplementary data.

• 2)

  Selection criteria

  Studies were eligible for inclusion if they fulfilled the criteria summarized in Table 1.

  Table 1.

  Selection criteria for studies assessing effectiveness and safety.

  Criteria category	Inclusion criteria
  Population	Individuals with chronic pain secondary to mild or moderate KO refractory to standard treatment and/or severe KO who are not candidates for surgery
  Intervention	GAE
  Comparator	Standard treatment, usual treatment, conservative management, drug therapy (without GAE) or non-intervention
  Outcome measure	Outcomes of interest included KO pain, overall function, HRQoL, need for pain medication, complications, and adverse events. For economic evaluations, the outcomes were: ICER, costs expressed in monetary units, and benefits expressed in QALY, LYG, monetary units, or any of the outcome measures included in the safety or effectiveness section
  Type of study	High-quality systematic reviews were included, as well as randomized controlled trials. In their absence, non-randomized clinical trials and observational comparative studies were consideredFull economic evaluations (either alongside primary studies or modelling based) were included: cost-benefit analysis, cost-utility analysis, cost-effectiveness analysis, cost-consequence analysis, and cost-minimization analysis
  Language	Studies published in Spanish or English
  Publication type	Only full original publications

  GAE: genicular artery embolization; HRQoL: health related quality of life; ICER: incremental cost-effectiveness ratio; KO: knee osteoarthritis; LYG: life years gained; QALY: quality-adjusted life years.
• 3)

  Study selection, data extraction process and assessment of risk of bias

  Two reviewers independently and in parallel performed the selection process, data extraction, and assessment of risk of bias. Titles and abstracts were initially assessed, followed by a full-text review for studies meeting the criteria. A standardized data extraction form was created in Microsoft Excel, including information on article identification, study design, participant characteristics, eligibility criteria, intervention and comparator details; assessed effectiveness and safety outcomes along with measurement timing; cost-effectiveness methodology and outcomes; study funding source, and disclosed conflicts of interest by the researchers.

  To assess the methodological quality of the identified systematic reviews, the AMSTAR-2 tool was selected.24 The risk of bias of the included primary studies was assessed using the Cochrane RoB 2 tool for parallel and crossover randomized controlled trials.25 The appraisal of methodological quality of economic evaluations was planned based on Drummond et al.’s criteria list.26

  Discrepancies between reviewers were resolved through discussion.

• 4)

  Publication bias assessment

  The assessment of publication bias was planned by creating a funnel plot and computing the Egger's regression test;27 however, the minimum number of studies necessary was not reached.

• 5)

  Synthesis of the evidence

  Study characteristics were summarized narratively and presented in summary tables. Meta-analyses were conducted for quantitative synthesis. Heterogeneity was assessed using forest plots, the χ2 statistical (p<0.01) and the Higgins I2 test. Initially, a fixed-effect model was used to evaluate the statistical heterogeneity among included studies (I2 statistics). In cases of high unexplained heterogeneity (I2>70%), meta-analyses was not performed, and results were reported narratively. All analyses were conducted using RevMan 5.4.28

  Potential confounders considered were baseline pain level, osteoarthritis severity, and type of embolic agents used. However, subgroup analyses were limited due to the small number of studies evaluated.

• 6)

  Certainty of evidence

  The certainty of the evidence for all outcomes was judged using the GRADE methodology, considering risk of bias, consistency, directness, precision and reporting bias.29 Certainty was assessed as high, moderate, low or very low. A summary of findings table was prepared to present the certainty of the evidence and the magnitude of the effects for the main comparison.30,31

A cost-analysis was conducted to compare the costs (in Euros of 2023) of two treatments for chronic pain secondary to KO: the use of GAE of the affected knee followed by standard treatment (evaluated strategy), and the standard treatment alone (comparator), which involves annual hyaluronic acid injections and prescription of physical exercise, in addition to radiofrequency sessions if the patient does not improve.

The target patients were those with chronic pain secondary to mild to moderate KO refractory to standard treatment or non-surgical candidates with severe KO. The analysis was performed from the perspective of the Spanish National Health System (NHS) with a one-year time horizon covering the entire treatment period in the evaluated strategy. No discount rate was applied due to the short-term horizon.

The incremental cost per patient and the annual cost for a hospital with a target population size of 40 patients (value reported by experts) were calculated. Additionally, we estimated the health improvement, in quality-adjusted life years (QALY), that GAE should generate per patient to be considered a cost-effective technology. This was estimated by solving the incremental effectiveness EA−EB from the incremental cost-effectiveness ratio equation (ICER=CA−CB/EA−EB, where CA−CB is the estimated incremental cost, and A and B represent the evaluated strategy and the comparator, respectively). The cost-effectiveness threshold estimated for Spain (€ 25 000 per QALY)32 was applied. The analysis was conducted using Microsoft Excel 2013.

We performed one-way deterministic and probabilistic (1000 second-order Monte-Carlo simulations) sensitivity analyses.

• 1)

  Parameters

Since both alternatives include the standard treatment, only the resource use and the corresponding costs of the GAE were considered, as the inclusion of shared costs would not affect the incremental cost.26

The analysis included those associated with the intervention (drugs, instruments, wound closure, healthcare personnel, operating room use, and post-intervention observation period) and the follow-up (Appendix B in Supplementary data).

The search in Epistemonikos yielded a total of 13 systematic reviews. Nine were initially considered potentially eligible after title and abstract screening according to the selection criteria; however, none were finally included because they did not attain a high-quality rating.4,6,13–16,33–35 The full quality assessment can be found in Appendix C in Supplementary data.

The search for primary studies retrieved 126 references after removing duplicates (Fig. 1). After the title and abstract screening, 12 publications were selected for full-text analysis. According to pre-established criteria, nine of these were excluded. Appendix D in Supplementary data shows the list of excluded references and the main reason for exclusion.

Figure 1.

Flow diagram of the selection process of effectiveness and safety studies.

(0.27MB).

Examination of the bibliographic listing of included studies and the Google Scholar search did not lead to any additional studies. No additional studies were identified through the alert service created using the Epistemonikos database. Therefore, the final selection consisted of two randomized controlled trials,36,37 reported in three publications.36–38

• 1)

  Characteristics of included studies

  The main characteristics of the included studies are summarized in Tables 2 and 3. For a detailed description of the characteristics and further details of the included studies, please refer to Appendix E and F in Supplementary data.

  Table 2.

  Selection criteria and baseline characteristics of participants in the included studies.

  First author, year	n	n woman (%)	KO grade	Inclusion criteria	Exclusion criteria	GAE group				Control Group
  						n	AgeMean±SD (range)	BMIMean±SD (range) / Median (IQR)*	Other characteristics Mean±SD (range) / Median (IQR)*	n	AgeMean±SD (range)	BMIMean±SD (range) / Median (IQR)*	Other characteristics Mean±SD (range) / Median (IQR)*
  Bagla, 202237	21	18 (85.7)	1-3 KL	1. >40 years2. KL grade1-3 findings on knee radiography3. ≥50/100 on the VAS for pain4. Pain refractory to 3 m of conservative therapies	1. Local infection2. Life expectancy of <6 m3. Severe atherosclerosis seen on prior imaging4. RA5. Infectious arthritis6. Prior knee replacement surgery7. INR of >2.58. Platelets less than 30,000/μL9. Iodinated contrast medium allergy resulting in anaphylaxis10. eGFR of <60 mL/min/1.73 m2	14	63.9±8.37(49-78)	30.8±8.14(16.9-43.8)	KL: 2.3±0.51(2-3)WOMAC: 64.9±17 (33-87)VAS: 81.3±12 (55-99)	7	62.9±7.13(49–71)	33.4±10.5(21.5-52.9)	KL: 2.3±0.76(1-3)WOMAC:70.9±13 (56-94)VAS: 78.9±10 (69-92)
  Landers, 202336	59	37 (62.7)	2 KL	1. 18-75 years2. KL grade 23. Knee pain ≥6 m4. Moderate to severe unilateral knee pain (≥3/10 knee pain on at least half the days in the preceding month according to an 11-point numeric scale)5. Resistant to conservative treatment for at least 6 m6. Willing, able and mentally competent to provide informed consent	1. Local infection2. Active malignancy3. RA or seronegative arthropathies4. Prior ipsilateral knee surgery, excluding arthroscopic surgery >6 m ago5. Ipsilateral knee intra-articular injection in past 6 m6. KL grade 3-47. Pregnant or attempting pregnancy during the study period8. Allergy to contrast media9. Reduced kidney function or failure 10. Body weight >200 kg11. Platelets <100×109/L12. INR >1.513. Approved for knee joint replacement surgery14. Moderate to severe pain in other lower limb joints15. History of allergy to carbapenem, or immediate/severe reaction to penicillin/cephalosporin antibiotic16. History of seizures or valproate use	29	61.1±8.0(NR)	30.3(27.8 -37.8)*	Knee pain (years): 1.5 (1-5)*	30	60.1±7.7(NR)	33.6(29.4-36.2)*	Knee pain (years): 1.0 (1-3)*

  BMI: body mass index; eGFR: estimated glomerular filtration rate; GAE: genicular artery embolization; IQR: interquartile range; INR: International Normalized Ratio; KO: knee osteoarthritis; KL: Kellgren-Lawrence; NR: not reported; RA: rheumatoid arthritis; SD: standard deviation; VAS: visual analogue scale; WOMAC: Western Ontario and McMaster Universities Osteoarthritis Index.

  *

  Median (IQR).
  Table 3.

  Characteristics of genicular artery embolization and control procedures in the included studies.

  First author, year	GAE group							Control group
  	Type of embolizer	Doses	Time (min)Mean±SD (range)	Anesthesiatype (details)	Sedationlevel (drugs)	Fluoroscopy time (min) Mean±SD (range)	Radiation dose (mGy) Mean±SD (range)	Type of control	Anesthesiatype (details)	Sedationlevel (drugs)
  Bagla, 202237	Absorbable particles (Opti Sphere; Teleflex, Minneapolis, Minnesota)	100-300micron	29.9±15.8(13-55)	NR	Moderate(midazolam+fentanyl)	6.70±4.96 (4.05-17.8)	17.9±5.52(8.8-23.8)	Sham procedure	NR	Moderate(midazolam+fentanyl)
  Landers, 202336	Imipenem and cilastatin sodium (IPM-CS, Primaxin; Merck & Co, USA)	0.5 g	NR(30-60)	Local (into the groin immediately superficial to the femoral artery)	Light (NR)	NR	NR	Sham procedure	Local (into the groin immediately superficial to the femoral artery)	Light (NR)

  GAE: genicular artery embolization; NR: not reported; SD: standard deviation.
• 2)

  Risk of bias in included studies

  The overall risk of bias in the included randomized controlled trials was considered low. The summary of the assessment can be found in Figure 2 and Appendix G in Supplementary data.

  
  

  Figure 2.

  Risk of bias in included studies.

  (0.35MB).
• 3)

  Certainty of evidence

  The overall quality of evidence was considered low. The evidence profile for GAE vs. standard treatment outcomes indicated moderate to low certainty of the evidence (Appendix H in Supplementary data).

• 4)

  Summary of results

  Results of all meta-analysis conducted are available in Appendix I in Supplementary data.

• •

  Pain (certainty of the evidence: low ⊕⊕⊖⊖/moderate ⊕⊕⊕⊖)

  The included randomized controlled trials examined the effect of GAE on pain at 1 month,36,37 6 months36, and 12 months.36 However, due to very high heterogeneity rates (I2=91%) in the 1-month pain analysis, pooled data are not presented.

  Preliminary findings reveal significant discrepancies in pain levels 1 month after GAE. There is low-quality evidence suggesting that GAE may have little to no effect according to the KOOS Pain subscale (1 study; 59 patients)36 or may result in a slight decrease in pain levels assessed with the visual analogue scale (VAS) (mean difference [MD]=50.1mm; 95% confidence interval [95%CI]: 29.0-72.3; 1 study; 21 patients).37 At 6 and 12 months of follow-up, moderate quality evidence suggests GAE probably results in little to no effect.

  In the Bagla et al.37 study, the response rates at 1 month were 79% (11/14) and 0% (0/7) for the GAE and sham arms, respectively. However, in the Landers et al.36 study, overall change in knee pain at 12 months indicated that 17 participants (58.6%) in the GAE group reported being moderately or much better, compared to 11 participants (37.9%) in the control group, though this difference was also not statistically significant.

• •

  Overall function (certainty of the evidence: moderate ⊕⊕⊕⊖)

  The included studies examined the impact of interventions (GAE vs. sham) on overall functional improvement at 1 month,36,37 6 months,36 and 12 months,36 using the WOMAC,37 KOOS Daily Living subscale, and KOOS Function in Sport and Recreation subscale.36 The effects in all these scales were pooled.

  Moderate-quality evidence suggests GAE likely has little to no effect on functional capacity at 1 month (SMD=−0.18; 95%CI: −0.62-0.27; I2=0%, 2 studies, 80 patients), 6 months (SMD=−0.17; 95%CI: −0.68-0.34, 1 study, 59 patients) and 12 months (SMD=0.07; 95%CI: −0.44- 0.58, 1 study, 59 patients) compared to standard or pharmacological treatment.

• •

  Health related quality of life (HRQoL) (certainty of the evidence: low ⊕⊕⊖⊖/moderate ⊕⊕⊕⊖)

  Only Landers et al.36 reported on the effect of interventions on HRQoL. Low-quality evidence suggests that GAE may have little to no effect on HRQoL assessed with the KOOS Quality of Life subscale at 1 month compared to standard treatment (1 study, 59 patients). GAE probably results in little to no difference in HRQoL at 6 and 12 months.

  However, low-quality evidence suggests that GAE may result in a slight improvement of HRQoL levels using the EQ-5D VAS at 6-month follow-up (MD=−10.00; 95%CI: −19.45 to −0.55; 1 study, 59 patients). Moderate-quality evidence suggests that GAE likely has little to no effect on HRQoL at 1 month and 12 months compared to standard treatment (1 study, 59 patients).

  No differences were observed in the rates of patients who did not present or presented slight problems with anxiety, discomfort, mobility, usual activities, or self-care assessed with EQ-5D.

• •

  Need for pain medication (certainty of the evidence: moderate ⊕⊕⊕⊖)

  Only Landers et al.36 reported changes in the need for pain medication. At 12 months, the GAE group had a lower proportion of participants taking analgesics (control 48%, intervention 24%). However, moderate-quality evidence suggests that GAE likely does not reduce or increase the need for pain medication (risk ratio: 0.52; 95%CI: 0.24-1.10; 1 study, 59 patients).

• •

  Adverse events and complications

  No major adverse events were reported in either of the two studies.36,37 Specifically, in the study by Landers et al.,36 it was reported that no evidence of osteonecrosis or ischemic complications were found on magnetic resonance imaging up to two years following the procedure. No differences were observed in minor adverse events.36,37

The electronic databases retrieved nine references, but after reading titles and abstracts, all of them were excluded (Appendix J in Supplementary data). Therefore, no economic evaluations focused on GAE and that meet the established inclusion criteria were identified.

Over a one-year horizon, the inclusion of GAE in the standard treatment generates an increment in costs of € 3432.37 per patient, from the NHS perspective (Table 4). Considering this incremental cost, a health improvement of 0.137 QALY per patient is required for GAE to be deemed cost-effective compared to the standard treatment. In addition, the estimated annual cost of implementing this technology would be € 137 294.86 for a general hospital treating a target population of 40 patients per year.

The one-way sensitivity analysis shows that the incremental cost can vary from € 2884.14, if the intervention duration is reduced from 1.5hours to 1 hour, to € 9722.02, if GAE is administered three times to the same patient within the time horizon. Meanwhile, the probabilistic analysis estimates an incremental cost of € 6741.84 per patient (95%CI: 2992.10-11,726.68), associated with the application of the GAE technique (Table 4).