Device Related Thrombosis and Bleeding in Pediatric Health Care: A Meta-analysis Free

We prospectively developed and registered a review protocol (PROSPERO registration number: CRD42021254922). The review has been reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses⁹  and the Meta-analysis of Observational Studies in Epidemiology reporting guidelines.¹⁰ 

A systematic search for published and unpublished cohort studies and clinical trials examining complications of indwelling invasive devices used in pediatric health care was initially conducted on May 25, 2021, and an updated search was conducted on June 16, 2022. An information specialist developed the Medical Subject Heading-informed search criteria (Supplemental Table 5 for references in online-only supplementary materials) and searched databases for published studies (Medline [via PubMed], CINAHL, Embase, Web of Science, Scopus, Cochrane CENTRAL), clinical trial registries (Clinicaltrials.gov, Australian New Zealand Clinical Trials Registry, International Clinical Trials Registry Platform), and unpublished study databases (ProQuest Dissertations and Theses, Open Access Theses and Dissertations, and MedNar). Reference lists of included articles were also screened. The results of all searches were entered into the Covidence software program¹¹  for screening. A large search was performed to identify a range of invasive device outcomes (infection, thrombosis, device performance), with the results separated into 3 separate meta-analyses for clarity.

To be included, primary studies had to include: (1) indwelling invasive devices (ie, a medical device penetrating inside the body, either through an orifice or skin for an extended period (ie, not only during procedures), used in pediatric health care,^12,13  (2) pediatric participants admitted to a hospital (postneonatal or maternal discharge to 18 years), (3) report device-related thrombosis or bleeding, and (4) using cohort (retrospective or prospective) or a clinical trial (with an established standard care arm) design. Studies with full-text data, available in English, and published from January 2011 to June 2022, were included. We excluded studies based in NICUs and home settings (because of separate health service systems and clinical teams).

Outcomes evaluated were device-related thrombosis (categorized as symptomatic, asymptomatic, and unspecified) and postinsertion major and minor bleeding (definition detailed in Table 1). These complication definitions were aligned with international guidelines.^14–16 

As per Meta-analysis of Observational Studies in Epidemiology recommendations,¹⁰  the eligibility criteria were applied to the systematic search results, and all identified references were screened independently by 2 reviewers using a 2-stage approach. The screening of titles and abstracts and the full-text review were performed by 2 authors using Covidence systematic review.¹¹  Any disagreement between these investigators was resolved by discussion with the senior author.

Two authors independently extracted relevant data, including author, publication year, country, study design, setting, number of patients and devices, device type, study method, the case of device-related thrombosis and bleeding complications, and device days. The extracted data were reviewed by a third investigator.

Each article was independently evaluated for methodological quality by 2 reviewers using the Joanna Briggs Institute Critical Appraisal Checklist for systematic reviews of prevalence and incidence.¹⁷  Any disagreement was resolved by a third reviewer. Each question was assessed as either a “yes,” “no,” “unclear,” or “not applicable” with the total number of points for each paper calculated. A quality score was assigned to each article, represented as a percentage, which was calculated by dividing the total number of points scored by a paper by the total number of questions for the given tool. Studies were not excluded based on their bias assessment.

Score confidence intervals (CI) with Freeman-Tukey double arcsine transformations were calculated for studies with dichotomous outcomes (complication or no complication), and Poisson confidence intervals and SEs were calculated for incidence rate (IR) outcomes. Pooled estimates were generated with random-effects meta-analysis and presented with a 95% CI. IR outcomes (continuous data) were pooled by using inverse variance with the DerSimonian and Laird method, per 1000 device days and 95% CI. Heterogeneity between studies was assessed using I-squared statistics, categorized as low (<25%), moderate (25% to 75%), or high (>75%).¹⁸  Outcomes with a single study were reported descriptively. Posthoc subgroup analyses by patient settings (intensive care, oncology and hematology, all pediatrics [across all hospital settings], renal, cardiac, neuro, and gastro-enterology-nutrition and hepatology) were conducted to examine differences in outcomes between varying risk patient populations and to investigate heterogeneity, which is reported in the main findings per each device. Sensitivity analyses were performed by the risk of bias (ROB) and study design. The ROB of the included studies was assessed using the Joanna Briggs Institute tool for prevalence studies (Supplemental Table 6).¹⁷  The defined questions were scored with 1 for “Yes” and 0 for “No” or “Unclear.” The total score of each article was calculated by the sum of its points. Based on this tool, studies were rated as low risk and high risk with scores 0 to 7 and 8 to 9, respectively. Statistical analysis was performed using Stata 15 (Stata Corp, College Station, TX), with a statistical significance of P value < .05.

Figure 1 illustrates the study selection, summarized with a Preferred Reporting Items for Systematic Reviews and Meta-Analyses study flow diagram. The initial search identified a total of 3951 potentially relevant studies. After screening the titles and abstracts, 744 full-text articles were assessed for eligibility. Eventually, 107 studies met the inclusion criteria of device-related thrombotic and bleeding complications and were included in the review (Table 2).

The summary of the included studies is shown in Table 2. A total of 57 360 devices and 1 701 559 device days were included. Most studies were retrospective (66 studies; 62%) or prospective (29 studies; 27%) cohort studies, based in North America (52 studies; 49%), involving children admitted across a pediatric hospital (31 studies; 29%), to the PICU (26 studies; 24%), or to an oncology and hematology unit (23 studies; 22%). The review found 9 invasive devices; the most commonly studied device was the central venous access device (CVAD; 71 studies; 66%), followed by ventricular assist devices (VAD; 15 studies; 14%). There were 46 studies (43%) that scored over 80% on quality scores (Supplemental Table 6), but common quality issues included unclear outcome definition, conditions not measured in a standard, reliable way for all patients, and unclear response rate or management.

Table 3 presents the pooled proportion and IR of thrombosis and bleeding categorized by device types. Proportions and incidence rates of thromboses and bleeding complications across device types by settings are summarized in Table 4.

In relation to the complications associated with CVADs, the pooled proportion of symptomatic venous thromboembolism (VTE) complications in CVADs was 4% (95% CI, 3–5; 23 studies; 25 352 devices) and IR 0.03 per 1000 device days (95% CI, 0.00–0.07; 10 studies; 813 221 device days), whereas the pooled proportion of asymptomatic VTE complications in CVADs was 10% (95% CI, 7–13; 42 studies; 25 416 devices) and IR 0.25 per 1000 device days (95% CI, 0.14–0.36; 17 studies; 732 457 device days). Both estimates had high heterogeneity and significant effect sizes for proportion and IR for asymptomatic VTE, but IR for symptomatic VTE had moderate heterogeneity. The pooled proportion of major bleeding was estimated at 4% (95% CI, 0–13; 3 studies; 1103 devices) and IR of 1.39 per 1000 device days (95% CI, 0.00–3.27; 1 study; 2166 device days), whereas the pooled proportion of minor bleeding was estimated at 3% (95% CI, 1–5; 11 studies; 3039 devices) with IR of 0.12 per 1000 device days (95% CI, 0.00–0.25; 6 studies; 221 257 device days). IR for minor bleeding had a moderate heterogeneity.

In the subgroup analyses, for patient settings of more than 2 studies, PICU had the highest symptomatic VTE proportion (6%, 95% CI, 3–9; 3 studies; 559 devices). For asymptomatic VTE, renal settings had the highest proportion (61%, 95% CI 54–68, 2 studies; 170 devices), and PICU settings had the highest IR (6.05 per 1000 devices days, 95% CI 1.67–10.44, 3 studies; 15 555 device days).

Only 1 study reported asymptomatic VTE with a proportion of 20% (95% CI 10–37; 30 devices). The pooled proportion of minor bleeding was 5% (95% CI 3–8; 2 studies; 296 devices). The studies were all from all pediatric settings.

There was only 1 study with an arterial catheter device for symptomatic VTE (3%, 95% CI 2–5; 615 devices) and 1 study of asymptomatic VTE (0%, 95% CI 0–2; 228 devices). The pooled proportion of minor bleeding was 14% (95% CI 5–25, 3 studies; 339 devices) from all pediatrics, PICU, and cardiac settings.

For unspecified thrombosis, extracorporeal membrane oxygenation (ECMO) had a pooled proportion of 19% (95% CI, 7–34; 5 studies; 353 devices). Pooled proportion for major bleeding was 67% (95% CI, 52–81; 3 studies; 269 devices), and 12% for minor bleeding (95% CI 6–21, 2 studies; 84 devices). The subgroup analysis showed that the PICU setting (23%, 95% CI 7–44, 3 studies, 330 devices) had a higher proportion of unspecified thrombosis compared with the cardiac setting (13%, 95% CI 1–31, 2 studies; 23 devices).

VADs had a pooled proportion of 33% (95% CI 13–57; 9 studies; 725 devices) and IR of 2.47 per 1000 device days (95% CI 0.69–4.25, 1 study; 3 647 device days) for unspecified thrombosis (Supplemental Fig 2). For major bleeding, VADs had an overall pooled proportion of 28% (95% CI 19–39; 14 studies; 1223 devices) and IR of 1.10 per 1000 device days (95% CI 0.00–2.35, 1 study; 3647 device days). The subgroup analysis yielded similar proportions for unspecified thrombosis and major bleeding across PICU and cardiac settings.

Nasogastric and gastrostomy tubes had a pooled proportion of 0% (95% CI 0–1, 2 studies; 1418 devices) for major bleeding and 23% for minor bleeding (95% CI 14–33; 2 studies; 82 devices). Major bleeding had only 1 setting (all pediatrics), and minor bleeding only had 1 study each for gastro-enterology-nutrition and hepatology and oncology and hematology.

Endotracheal and tracheostomy devices had a pooled proportion of 0% for minor bleeding (95% CI; 0–2; 240 devices) across 3 studies conducted in a PICU setting.

Peritoneal dialysis catheter had no major bleeding (95% CI 0–10) reported in a single study from a renal setting.

External ventricular drainage had a pooled proportion of 18% for major bleeding (95% CI 11–28) from a single study with 73 devices.

In a sensitivity analysis, pooled proportions for symptomatic VTE in CVADs were lower for higher ROB studies (3%, 95% CI 2–5; 14 studies; 11 526 devices vs 6%, 95% CI 3–11; 9 studies; 13 826 devices) with the test for heterogeneity between subgroups not significant (Supplemental Table 7). The IR was also lower in higher ROB studies (0.04, 95% CI 1–6; 6 studies; 731 810 device days vs 0.07, 95% CI 0.00–0.26; 3 studies; 79 245 device days). The heterogeneity was moderate for higher ROB studies, and the heterogeneity between subgroups was not significant. For asymptomatic VTE in CVADs, the higher ROB studies had higher pooled proportions compared with lower ROB studies and also had higher pooled IR. In general, the prospective study design had a higher pooled proportion for asymptomatic CVAD, unspecified thrombosis, and major and minor bleeding. However, IR was higher in retrospective study design in symptomatic and asymptomatic VTEs in CVADs (Supplemental Table 8).

Our meta-analysis, involving 107 studies (57 360 devices, 1 701 559 device days), has demonstrated that children hospitalized across the tertiary and regional health sectors remain at considerable risk for thrombosis and bleeding complications associated with their invasive devices.

The findings of this review are in accordance with previous studies that have reported high rates of thrombotic complications associated with CVADs, including pooled 4% symptomatic (IR 0.03 per 1000 device days) and 10% pooled asymptomatic (IR was 0.25 per 1000 device days) venous thromboembolism. However, despite their relative commonality, the sequelae of asymptomatic thrombosis in CVADs are unclear for general cohorts, with single-site studies demonstrating these thromboses, when left untreated, had a minimal impact on health experiences in medically complex and high-risk patients.⁴  To inform the necessity of routine surveillance and antithrombotic prevention for device-related thrombosis, multicenter studies with blinded radiologic assessments are needed to investigate the natural history of asymptomatic VTEs.

The high rates of major bleeding associated with both VAD (28%; 14 studies, 1223 devices) and ECMO (67%; 3 studies, 269 devices) are concerning. These rates may be because of underlying pathologies, iatrogenic coagulopathy, and anticoagulation regimens. Ironically, many anticoagulant therapies are commenced to prevent or treat device-related thromboses.^19,20  Although varying in severity, this type of harm can change their health trajectory, causing immediate morbidity, mortality, and delayed treatment.^21,22 

ECMO cannula were also associated with high rates of thrombosis; however, most studies did not report clear outcome definitions, influencing accuracy. The ELSO (Extracorporeal Life Support Organization) registry is an international database that collects and analyzes data on patients who received ECMO support. The data definition on thrombosis is only defined as “cannula exchanged primarily because of clot burden within the cannula” or “circuit component (eg, pigtails, connectors, bridge, arterial, or venous tubing) requiring change because of clot formation or mechanical failure of the component, not equipment,”²³  which ignores the thrombosis that do not result in circuit changes and lead to a variety of thrombosis outcome definitions. Therefore, it is crucial to establish a universal consensus on outcome definitions, such as VTE definition in CVADs,²⁴  to ensure accurate and standardized reporting of complications.

Although this result is limited by the comparatively small sample size and use of multiple concurrent devices, the overall picture of high risk in a highly vulnerable patient group is clear. Thrombosis and bleeding complications can severely affect patient outcomes and should be a priority target for significant technology and innovation.^25,26  There are key practices and technologies that may reduce device-related complications in pediatric health care. Many of these, such as prophylactic anticoagulation,¹⁶  antithrombotic catheter lock therapies,²⁷  antithrombogenic or antimicrobial materials,^16,28,29  antithrombotic catheter lock therapies,²⁷  antithrombogenic or antimicrobial materials,^28,29  and topical hemostatic agents,³⁰  show potential clinical benefit but are yet to be rigorously evaluated for clinical effectiveness and pediatric practice contexts.

Our meta-analysis has also highlighted several inequities. In particular, most studies were conducted in highly technical, resource-rich settings (eg, North America, PICUs). The rates of thrombotic and bleeding complications associated with invasive devices in other settings are unclear, especially when considering social determinants of health influencing premorbid conditions and treatment options.^31,32  The study also draws attention to multiple understudied devices and outcomes, often not included in traditional, infection-related surveillance reports.^31–33  The study also draws attention to multiple understudied devices and outcomes, often not included in traditional, infection-related surveillance reports.³³ 

The review has limitations. Firstly, a consistent definition was absent across all device types, and the ambiguous outcomes have been removed. Some studies lacked clarification on whether the complications were because of the device or because of the treatment regimen the patient was on (ie, thrombosis in ECMO or VAD), and these studies were removed from the meta-analysis. Additionally, this meta-analysis included only the inpatient studies; hence, it may not be generalizable to out of hospital and community settings. This removal of these studies may have resulted in estimate imprecision and generalizability. Secondly, the estimate could also be underestimated for some devices because of lower publication frequency (eg, peripheral intravenous catheters) and relying on author-defined outcome events. It is crucial to establish and apply universally agreed definitions for all common device-related complications, like those that have been developed for infections, for studies, and clinicians to be able to benchmark their findings.²⁴  A future study incorporating individual participant data meta-analysis method³⁴  or weighted analysis using utilization rate could potentially quantify a more reliable estimate. Thirdly, overall, the meta-analysis had high heterogeneity across studies and within subgroups. This is expected because of the heterogeneous nature of patients included in the study,³⁵  which is demonstrated in the study population in Table 1 and was the rationale for the subgroup testing. Lastly, the majority of included studies were retrospective cohort studies, which are prone to bias and confounding. Despite the limitations, this meta-analysis can provide an opportunity to target innovation areas, highlighting the areas required for further investigation.

This systematic review provides valuable insights into the rates of thrombotic and bleeding complications associated with individual invasive devices in pediatric patients. The findings highlight the need for improved outcome definition and surveillance strategies of these devices in clinical practice. Further research is needed to better understand the factors contributing to these complications and to develop effective prevention and management strategies, potentially leveraging existing technologies used in other medical devices.

We thank the health librarian (Mr David Honeyman) for their assistance in putting together our search strategy; and Dr Xu and Ms Lions for assisting us with the data collection.