Diagnostic accuracy of ultrasound for localising peripherally inserted central catheter tips in infants in the neonatal intensive care unit: a systematic review and meta-analysis

Peripherally inserted central catheters (PICC) are ubiquitous in the neonatal intensive care unit (NICU) and are used to facilitate administration of total parenteral nutrition in low-birth-weight infants, and long-term administration of intravenous antibiotics [1]. The establishment of long-term venous access is also advantageous because it averts the need for repetitive catheterisation, thereby reducing pain and risk of infection for the infant whilst limiting excessive stimuli and extended time away from dedicated nursing care [2].

The use of PICCs in clinical practice is associated with various complications. Most of these complications are associated with malpositioning or migration of the catheter tip [3‐5]. If the catheter tip is incorrectly placed, or migrates following insertion, pericardial effusion, thrombosis, infection, perforation, arrhythmia, cardiac tamponade, heart valve damage and patient discomfort can occur [6, 7].

Accurate PICC tip positioning is essential to minimise the risk of complications. The international standard for upper body inserted PICC tip position is at the caval–atrial junction at the lower third of the superior vena cava [8]. For PICCs inserted via the lower limb, the tip should be within the upper inferior vena cava between the 9th and 11th thoracic vertebrae [9, 10]. PICC tip localisation and confirmation can be achieved through myriad diagnostic imaging modalities. Verification of the catheter tip in infants is typically undertaken using the reference standard for PICC confirmation, an anteroposterior chest radiograph for upper-body-deployed PICCs and a combined chest–abdominal radiograph for PICCs deployed via the lower limb [11, 12].

There is evidence to suggest that chest radiography alone is not entirely accurate at identifying intra-atrial tip position [13, 14]. Additionally, it is often difficult to position the neonate for an optimum radiographic image without geometric or movement artefact. Chest radiography also provides only a static image of PICC position. Arm movements have been shown to influence the position of the PICC, so a radiograph captures the tip position in respect to the limb position of the neonate at the time of exposure and might not be entirely representative of the long-term position [15, 16].

Repositioning of misplaced PICC lines requires supplementary radiographs to re-assess tip position. This exposes the infant to additional ionising radiation repeatedly, until the catheter tip is in the appropriate position. Neonatal and paediatric populations are considerably more susceptible to ionising radiation because the rate at which their cells undergo mitosis is more rapid than that observed in adult populations [17]. Increased radiosensitivity, greater mitotic activity and a protracted period for consequences to manifest mean that the risk of radiation-induced cancer per unit of dose is 2–3 times higher for preterm infants than the average population [18]. Reduction in the risk of radiation-associated comorbidities can be achieved by limiting unnecessary radiographs by employing alternative methods to achieve similar diagnostic information or conclusions [19‐21].

Bedside radiographic imaging in the NICU can be challenging for radiographers because the environment can make it impractical to comply with standard examination protocols. Neonates are unable to control their movements, making positioning accuracy difficult to accomplish. It is often the case that various lines, tubes and devices overlie the area of interest, interfering with image quality and diagnosis [22].

Neonatal intensive care unit staff are often required to be within 2 m of the X-ray source to hold the neonate, with their hands sometimes inadvertently being caught within the primary X-ray beam. A study by Russell et al. [23] showed that 15–40% of mobile radiographic examinations conducted in the NICU had at least one adult finger visible on the resultant image. In addition to direct radiation exposure, staff and patients in the NICU may also be exposed to secondary scatter radiation. Employing alternative imaging modalities that avoid the use of ionising radiation, such as US, is becoming increasingly popular, particularly for routine checks of catheter or tube positioning [24‐26].

Technological advancements have expanded the role that US can play in the critical care setting through improved image quality and accessibility. Advantages of US include real-time assessment and limited handling of critically ill infants. Evidence shows that minimal training is required to reliably perform quantitative US [27]. Additionally, clinical reporting can be performed by the sonographer at the time of scanning, which can reduce report turnaround time. However, US does carry a higher infection control risk and therefore good basic hygiene standards are essential [28].

Previous systematic reviews [29, 30] have shown the success of US in localising central venous catheters in comparison to the reference standard (radiography), but these studies have not investigated the accuracy of US in a neonatal population. The aim of this systematic review was to identify, critically appraise and assess the sensitivity and specificity of US in confirming PICC tip position in infants on the NICU.

Using the Cochrane protocol, we conducted and report a systematic review (CRD42020223684) in adherence with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) criteria/guidelines [31].

The PRISMA flowchart (Fig. 1) summarises the results of the search and review process. The search strategy yielded 120 studies following the removal of duplicates. We screened the abstracts and titles of these 120 papers and completed a full-text review of 15 studies. Eight papers were deemed eligible for inclusion. Potential studies for inclusion in this systematic review were appraised using the modified QUADAS-2 tool [33] (Table 2; [5, 35‐41]). All eight studies were included because they had an overall low risk of bias.

Details of the study characteristics for all eight papers are detailed in Table 3 [5, 35‐41]. Across the eight studies, 421 infants received a PICC line and 341 US scans were conducted. All eight studies reported the gestation for those neonates included, with a mean gestation for the eight studies of 38.6 (range 25.4 to 35.0) weeks. The mean birth weight recorded for included neonates was 1,581.1 (range 525.0 to 3,302.0) grams. All infants included in the study were in-patients in the NICU.

Five studies detailed the PICC insertion site, with 215 (59.9%) in the upper limb and 103 (28.7%) in the lower limb; 38 (10.6%) were superficial temporally located and three (0.8%) were auricle. Three studies (n = 64) did not report the exact insertion site of the PICC lines, but each of the three studies reported inclusion of both upper- and lower-extremity PICC line insertions [3, 21, 35].

All included studies reported diagnostic accuracy of US in relation to the reference standard. All included studies used radiography as the reference standard. The reference standard in six studies was solely chest radiography. In two studies the reference standard was both supine chest and abdominal radiography [36, 37].

Six studies provided details regarding the qualification and level of experience of the US operator involved (Table 3). The frequencies, brand and type of the US probes and machines used in the eight studies are reported in Table 4 [5, 35‐41].

All included studies acquired the reference standard image using a mobile radiography machine, although they did not specify the machine manufacture or detector type. All studies had a radiologist report the radiographs, except one, in which a neonatologist reported the radiographic images [37].

Four studies documented that the reference test was reported incorrectly. These erroneous radiographic reports were ultimately corrected following the US examination, which demonstrated a more accurate tip position [36, 38‐40]. Radiographic misinterpretations in these four studies were a mixture of false reports of malposition and false reports of optimum position. Diagnostic accuracy measurements from included studies are summarised in Table 5 [5, 35‐41].

The estimated sensitivity and specificity for US were 95.2% (95% CI: 91.9–97.4%) and 71.4% (CI: 59.4–81.6%), respectively. The pooled positive likelihood ratio was 3.3 (95% CI: 2.3–4.8) and the negative likelihood ratio was 0.07 (95% CI: 0.04–0.12). The positive predictive value and the negative predictive value were 92.8% (95% CI: 89.9–94.9) and 6.7% (0.04–12.0), respectively. Figure 2 shows the forest plot of the overall odds ratios. The ROC curve for the pooled diagnostic accuracy of US across all eight studies indicated an acceptable overall accuracy (area under the curve [AUC] = 0.83) (Fig. 3).

Mobile radiographic imaging is the modality employed in the NICU for confirming PICC tip location. An imaging pathway free from ionising radiation, such as US, is a contemporary and novel approach to confirming the position of the PICC line tip in a neonatal population. By acknowledging the limitations of conventional radiography, it is reasonable to suggest that questions might soon be raised regarding the preferred primary imaging approach. This systematic review considers whether US is a viable alternative to mobile radiography in the localisation of the PICC line tip in a neonatal population.

Eight studies were eligible for inclusion in this review and meta-analysis. We calculated a pooled sensitivity of 95.2% (95% CI: 91.9–97.4%). The specificity values across all eight studies were variable because of misinterpretations made by the reference standard, lowering the pooled specificity value to 71.4% (95% CI: 59.4–81.6%). These results suggest US to be an excellent test for the localisation of the correctly positioned PICC line tip; however, it is less than perfect for identifying a misplaced or migrated PICC tip.

The low specificity can be explained by referencing the studies by Jain et al. [39] and Shabeer et al. [40]. Jain et al. [39] reported sensitivity and specificity values of 60% (95% CI 41.6–81.1) and 58% (95% 34.1–73.5%), respectively. Radiography inaccurately interpreted 4/11 misplaced PICC tips as appropriately placed, evidently impacting the false-negative rate, which had a negative impact on the sensitivity value. Furthermore, radiography incorrectly indicated that five of the correctly positioned PICC line tips were malpositioned, subsequently elevating the false-positive rate and negatively affecting the specificity. The study by Shabeer et al. [40] had a notably lower sensitivity value of 47.8% (95% CI 35.8–60.1%) in comparison to the other studies included in the analysis. The specificity value reported by Shabeer et al. was similar to that reported in the other studies (82.4% [95% CI 61.7–93.1%]). The limited specificity of US in detecting the PICC line tips in the study by Shabeer et al. [40] was also attributed to an imperfect reference standard. Chan et al. [42] acknowledged the variability amongst radiologists regarding the cavo-atrial junction, which would explain the misjudgements made on the reports in the studies by Jain et al. [39] and Shabeer et al. [40]. Taking the limitations of the reference standard into consideration, had both these studies been omitted from the meta-analysis, the pooled sensitivity and specificity values for the remaining seven studies would have been 97.5% and 91.4%, respectively.

The pervasiveness of PICC tip malposition within our systematic review aligns with that reported in the published literature in respect to the general population (10–53%) [43, 44]. Repositioning of these malpositioned or migrated catheter tips can be achieved under US guidance, avoiding the need for repeated radiographic imaging to reassess catheter tip location. This advantage of US has the potential to negate additional radiation exposure to the neonate [2].

Whilst our systematic review investigates the pooled diagnostic accuracy measurements of US as a confirmatory method for PICC line tip position in the neonatal population, our results align with recent publications exploring the accuracy of US for various other catheter and tube types. In 2017, a meta-analysis by Ablordeppey et al. [30] reported a pooled sensitivity of 82% (77–86%) and specificity of 98% (97–99%) of US as a confirmatory tool for central venous catheters. Similarly, a systematic review and meta-analysis published by Smit et al. [29] showed US to be an excellent alternative for the localisation of misplaced or migrated central venous catheter tips, reporting a pooled sensitivity and specificity of 68.2% (54.4–79.4%) and 98.9% (97.8–99.5%), respectively. The low sensitivity value was explained by an incorrect reference standard, further encouraging investigation into the accuracy of the reference standard chest radiography.

Our review extracted data in relation to the time taken to confirm PICC location using US. Four of the eight studies did not provide any data for the time taken to achieve PICC confirmation using US; however, all four of these studies reported that the US was faster than the reference standard. The remaining four studies reported time values. Three reported a scan time of less than 15 min [35, 39, 41]. The fourth study recorded a mean scan time of 60 min, which was considerably less than the time taken for conventional radiography (136 min) [40]. This element of the data extraction confirmed that US is a faster imaging modality than radiography for confirming catheter tip position in the neonatal population. In addition to this, US provides a real-time assessment, negating the time traditionally required for radiology reporting of the radiographic images.

Ultrasound quality is highly operator-dependent, making it necessary that US providers are appropriately trained. The lack of implementation of US as a contemporary imaging modality within the NICU setting is often a result of the lack of availability of appropriately qualified US staff [45]. The level of training of US operators varied across all eight studies included in our review. Six of the eight studies reported the qualification and training of the US operators. Despite varying levels of operator experience across all eight studies, diagnostic accuracy measurements did not vary significantly.

Our systematic review and meta-analysis provide evidence that US is sensitive and specific for confirming PICC line placement in the NICU. Our analysis estimated a pooled sensitivity of 95.2% (95% CI: 91.9–97.4%) and a pooled specificity of 71.4% (95% CI: 59.4–81.6%). In cases where US fails to locate misplaced PICC tips, radiography should be performed.