The systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines [37]. A comprehensive literature search of the databases PubMed, Ovid MEDLINE, and Ovid EMBASE was designed and conducted by an experienced librarian with input from the authors. The search duration was 2 months. The key words “sclerotherapy,” “vascular malformations,” “venous malformations,” “arteriovenous malformation,” “hemangioma,” “lymphatic malformation, “head,” “neck,” “facial,” “oropharyngeal,” and “orbital” were used in “AND” and “OR” combinations. The search was limited to articles published from 2000 to 2018. Inclusion criteria were the following: 1) English or Italian language, 2) case series reporting greater than 5 patients, 3) studies reporting image guided percutaneous sclerotherapy, 4) studies reporting exclusively face, head and neck VMs or subdividing outcomes and complications by anatomical region, and 5) studies classifying VMs appropriately using the International Society for the Study of Vascular Anomalies. Exclusion criteria were: 1) case series reporting fewer than 5 patients, 2) case reports, 3) vascular malformation not of the head and/or neck region (e.g., sclerotherapy for varicose veins in legs), and 4) studies not classifying lesions according to the International Society for the Study of Vascular Anomalies criteria. International Society for the Study of Vascular Anomalies criteria on imaging and clinical exam for VMs on imaging include 1) septated lobulated T2 hyperintense and T1 hypointense mass without mass effect, 2) phleboliths which are characteristically hypointense on T1/T2, 3) presence of fluid-fluid levels, 4) no flow voids on spin echo sequences, 5) the lesion infiltrates tissue planes, 6) no arterial or early venous enhancement, and 7) diffuse enhancement on delayed images.41 On clinical exam, VMs appear as faint blue, soft and easily compressible non-pulsatile masses. The lesions characteristically enlarged with Valsalva maneuver and in dependent positions and decompress with local compression.

In studies with overlapping patient populations written by the same author/institution, we only included the largest or most complete dataset. In cases where outcomes were separated out by the type of sclerotherapy agent used, we abstracted outcomes separately for each agent in order to perform our subgroup analyses. Two authors determined inclusion and exclusion criteria for the studies in the literature search with differences resolved by the senior author.

For each study, we extracted the following baseline information: number of patients, mean or median age and gender, number of malformations treated, location of malformations, sclerosing agent and its mean volume used, mean number of treatment sessions, and mean length of radiographic and clinical follow-up. The primary outcome of this study is the efficacy of sclerotherapy which includes complete cure of the vascular malformation (resolution of the VM on physical exam), partial cure of the vascular malformation (partial decrease in VM size), lack of benefit following sclerotherapy, improvement in quality of life (QoL), and patient satisfaction. Secondary outcomes are adverse events after sclerotherapy, including respiratory complications, skin necrosis/scars, any permanent morbidity/mortality, local temporary complications). Permanent morbidity and mortality were defined as mortality or any permanent neurological deficit. Local temporary complications included erythema, swelling, and pain.

For our subgroup analysis by sclerotherapy agent, we separated outcomes by agent. We were able to abstract data for the following individual agents: bleomycin, ethanol, sodium tetradecyl sulfate (STS), ethanolamine and pingamycin.

We modified the Newcastle-Ottawa Quality Assessment Scale to assess the methodologic quality of the studies included in this meta-analysis. This tool is designed for use in comparative studies; however, because the studies did not include a control group, we assessed study risk of bias based on selected items from the tool, focusing on the following questions: 1) Did the study include all patients or consecutive patients versus a selected sample?, 2) Was the study retrospective or prospective?, 3) Was clinical follow-up satisfactory, thus allowing ascertainment of all outcomes?, 4) Were outcomes clearly reported?, and 5) Were there clearly defined inclusion and exclusion criteria?

We estimated from each cohort the cumulative prevalence and 95% confidence interval (CI) for each outcome. Event rates were pooled across studies with a random-effects meta-analysis. Heterogeneity across studies was evaluated using the I² statistic. An I² value of >50% suggests substantial heterogeneity. We also extracted a 2×2 table to calculate P values for the comparisons among the results. For the purpose of statistical comparisons we chose bleomycin sclerotherapy as the reference group, since it is the sclerosing agent most commonly used in the USA. Meta-regression was not used in this study. Statistical analyses were performed using OpenMeta[Analyst] (http://www.cebm.brown.edu/openmeta/; Biostat, Englewood, NJ, USA).

The initial literature search yielded 1,211 articles. On review of the abstracts and titles, we excluded 1,126 articles. Eightyfive articles were selected for full-text screening, of which 37 met inclusion criteria [1-37]. The remaining 48 articles were excluded for reasons including 1) failure to separate outcomes by anatomic location (9 articles), 2) inclusion of lymphatic or venolymphatic malformations rather than pure VMs (14 articles), 3) use of confusing or unclear terminology making it difficult to ascertain whether lesions were VMs or hemangiomas (12 articles), and 4) mixture of VMs, AVMs and lymphatic malformations (13 articles). All studies included in the analysis had at least one or more outcome measure available for one or more of the patients groups analyzed. Fig. 1 shows the flow chart according to the PRISMA statement [37].

These 37 studies included a total of 2,067 patients. The smallest study included 10 patients and the largest included 358 patients. Mean age was 24.9 years. There was a female predilection (1:1.2). The mean number of malformations per patient was 1.08 and they were all located in the head and/or neck region. The highest number of treated malformations per study was 358, while the least was 10. The mean number of treatment sessions per patient was 2.4. The mean length of radiographic and clinical follow-up from the time of the first treatment was 16.61 months and 18.04 months respectively.

Most included studies used a single sclerosing agent for each vascular malformation, while the remainder used a combination of them. Five studies reported outcomes of bleomycin sclerotherapy, 10 studies reported outcomes of ethanol sclerotherapy, 7 studies reported outcomes of STS sclerotherapy, 4 studies reported outcomes of ethanolamine sclerotherapy, and 4 studies reported outcomes of pingamycin sclerotherapy. In 11 studies either multiple agents were used and we could not separate outcomes by agent or other sclerosing agent including OK-432 were used. A summary of included studies is provided in Table 1.

Overall complete cure rates were reported in 1,736 patients. The overall rate of complete cure following percutaneous sclerotherapy with any agent was 64.7% (95% CI, 57.4–72.0%). STS had the lowest complete cure rate at 55.5% (95% CI, 36.1–74.9%) while pingyangmycin had the highest cure rate at 82.9% (95% CI, 71.1–94.7%). Fig. 2 shows the complete cure rates forest plot. Overall partial cure rates were reported in 1,703 patients. The overall rate of partial cure following percutaneous sclerotherapy with any agent was 28.0% (95% CI, 22.1–34.0%). Partial cure rates ranged from 16.2% (95% CI, 4.7–27.6%) for pingyangmycin to 35.3% (95% CI, 19.1–51.5%) with ethanol. The overall rate of no benefit following percutaneous sclerotherapy was reported in 1,736 patients and was 4.5% (95% CI, 3.0–6.1%). STS sclerotherapy had the highest rate of no benefit (14.9%; 95% CI, 4.3–25.4%) while pingyangmycin had the lowest rate of no benefit (0.6%; 95% CI, 0.0–1.6%). Patient satisfaction was reported in 315 patients. Overall patient satisfaction rates were 91.0% (95% CI, 86.1–95.9%). Patient satisfaction rates ranged from 72.8% (95% CI, 63.6–81.9%) with STS to 96.0% (95% CI, 92.5–99.6%) with ethanol. Improvement in QoL was reported in 243 patients. Overall QoL improvement was 78.9% (95% CI, 67.0–90.8%) ranging from 46.7% (95% CI, 22.4–71.0%) for STS to 98.1% (95% CI, 94.5–100%) with ethanolamine.

Overall permanent morbidity and mortality rates were reported in 1,303 patients. The overall rate of permanent morbidity and mortality was 0.8% (95% CI, 0.3–1.3%). The lowest rate was reported with STS, and pingamycin (0.5%, 0.8%) and the highest rate reported was with bleomycin (2.2%; 95% CI, 0.0–5.4%). Fig. 3 shows the permanent morbidity and mortality rates forest plot. Local temporary complication rates were reported in 1,312 patients. The overall rate of local temporary complications was 41.8% (95% CI, 27.0–56.5%). Local temporary complications were highest with ethanolamine (51.0%; 95% CI, 24.7–26.6%) and were lowest with bleomycin (27.0%; 95% CI, 5.4–59.4%). Pulmonary complication rates were reported in 1,222 patients. The overall rate was 0.6% (95% CI, 0.2–1.0%). Pulmonary complication rate was lowest for pingamycin (0.8%; 95% CI, 0.0–2.3%), and was highest for STS (2.4%; 95% CI, 0.5–5.3%). Fig. 4 shows the pulmonary complication rates forest plot. Skin necrosis and scarring rates were reported in 1,273 patients and the overall rate was 1.5% (95% CI, 0.8–2.1%). Rates were highest with bleomycin (2.3%; 95% CI, 1.0–5.7%) and lowest with pingamycin (0.8%; 95% CI, 0.0–2.3%). Fig. 5 shows the skin necrosis and scarring rates forest plot. The efficacy and safety outcomes are summarized in Tables 2 and 3.

I² values were >50% indicating substantial heterogeneity for the following outcomes: lack of benefit, improvement in QoL, local temporary complication, overall cure, partial cure, patient satisfaction and systemic side effects. I² values were <50% indicating lack of substantial heterogeneity for the following outcomes: permanent morbidity and mortality, pulmonary complications and skin necrosis/scarring.

Our study has limitations. This meta-analysis was based primarily off of single-center case series and thus has limitations inherent to single center retrospective studies. There was high risk of bias in 21 series. While we were able to perform subgroup analyses based on sclerotherpeutic agent used, we were unable to perform more granular analyses stratifying outcomes by lesion location. Follow-up in the included studies is limited. Many of our outcomes are subjective and volumetric analyses documenting lesion improvement were not available for a vast majority of the included studies. Nonetheless, our study provides helpful information for both patients and providers who are considering percutaneous sclerotherapy for treatment of VMs and provides guidance for future areas of investigation.