Ultrasound-Assisted Lumbar Punctures in Children: An Updated Systematic Review With Meta-Analysis Free

This systematic review and meta-analysis followed the guidelines from the Cochrane Handbook for Systematic Reviews of Interventions¹⁶  and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.¹⁷ 

Study protocol was registered at PROSPERO (CRD42022372806).

We included randomized controlled trials (RCT) comparing the effectiveness of ultrasound-assisted LP with standard LP procedure. Only studies conducted in children were included. We excluded studies conducted in adults, non-RCT studies and conferences abstracts.

PubMed (Medline), Embase (via Elsevier), and the Cochrane Central Register of Controlled Trials (CENTRAL) databases were searched for relevant studies up to November 2022. The ClinicalTrials.gov and ClinicalTrialsRegister.eu websites were also searched for registered but not published RCTs. Two reviewers (A.C., M.K.), using Endnote ×9.3.3 independently, conducted the search, without any language restrictions. For search strategy please see Supplemental Tables 2–6.

The title of each identified study, with the applied search strategy and the abstracts of relevant articles, were screened. The deduplication process was done using Endnote ×9.3.3 software. Full texts were retrieved for each study potentially relevant for inclusion. Eligibility of the articles as well as the whole process of study identification and screening were assessed independently by 2 authors (A.C. and M.K.). In case of a disagreement, it was resolved by the discussion process.

Using a data extraction form, 2 reviewers (A.C. and M.K.) independently extracted information from each included study. The extracted data included: the last name of the first author, publication year, study country, study population size, study location (eg, emergency department [ED], operating room, ICU), mean age of study patients, sex of study patients, mean BMI, type of ultrasound transducer used, sonographer specialty.

Our primary outcome was the first attempted LP success rate regardless of the definition used by the authors. Secondary outcomes included: first attempted LP success rate according to RBC count, rate of traumatic LP, and mean time of the LP.

Risk of bias was assessed using Cochrane Collaboration tool.¹⁸  Two reviewers (A.C. and M.K.) independently performed the assessment, with disagreements resolved by discussion.

The dichotomous outcomes, the results for individual studies, and pooled statistics were reported as: risk difference (RD), odds ratio (OR), and risk ratio (RR) between the ultrasound and control groups with 95% confidence intervals (95% CI) using Mantel-Haenszel method. The continuous outcomes were reported as the mean difference between the ultrasound and control groups with 95% CI. We decided to choose RD, OR, and RR because they were used in previously published reviews, which makes it easier to compare the results. All analyses were based on the random effects model. The data were analyzed using Review Manager (RevMan Version 5.4.1 Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). All missing means and SDs were estimated using the formula recommended by Hozo et al.¹⁹ 

Heterogeneity was quantified by I², which can be interpreted as the percentage of the total variation between studies that is attributable to heterogeneity rather than to chance.

A funnel plot and Egger’s test with an α level of .05 were used to assess for publication bias for the primary outcome measure.²⁰  Despite that we included less than the recommended amount of 10 studies, we decided to publish the Funnel plots for the primary outcome.

For a flow diagram documenting the identification process for eligible trials, please see Supplemental Fig 3. Characteristics of the included RCTs are presented in Table 1, and characteristics of the excluded trials are presented in Supplemental Table 5.

We included 7 studies, published between 2014 and 2021, enrolling 618 patients.^21–27 

Four studies were conducted in the United States,^21,22,23,27  1 was conducted in Korea,²⁶  1 in Turkey,²⁴  and 1 in Canada.²⁵  The sample size in all included trials ranged from 26 to 166 participants. All studies were performed in the ED settings, although 1 study also included pediatric floor patients.²¹  The transducer type was not described in 2 studies.^26,27  Ultrasound was performed by an emergency physician in 6 studies and by a radiologist in 1 study.²⁴  Ultrasound was performed by attending physicians in 3 studies^21,24,27  and resident physicians in 1 study.²⁶  The remaining studies used a mix of providers or did not describe the sonographer experience level.^22,23,25  For the specific description of the POCUS procedure, as well as the description of the type of training received by the persons performing the ultrasound, please see Supplemental Table 6.

Most of the included studies were at risk for bias for at least 1 of the domains (see Supplemental Fig 4) None of the studies were at high risk for any domain. Randomization was not adequately described in 1 study.²⁶  Allocation concealment domain was at an unclear level in 4 RCTs because of limited description.^21,22,24,26  Blinding of the personnel and outcome assessment was unclear in almost all studies included. All studies were at low risk for the remaining domains: incomplete outcome data, selective reporting, and other bias. Overall, all the included studies were of moderate quality.

For our primary endpoint (success rate of first attempt LP regardless of the definition used by the authors), regardless of the type of statistics used, the heterogeneity was borderline significant (I2 ≥ 50%). I² for RD was 53%, for OR 45%, and it was 64% for RR. There was no evidence of publication bias for this outcome (for the funnel plot see Supplemental Fig 5). For other outcomes and subgroup analyses, publication bias was not formally assessed using a funnel plot because of the small number of studies included in the analyses.

We included 7 RCTs^21–27  with 618 participants, regardless of the definition of success rate used by the authors. Overall, ultrasound-assisted LP was successful in 224 of 309 patients (72.0%) and standard LP was successful in 184 of 309 patients (59.6%). The calculated RD was: 13.0% (95% CI: 3% to 23%) I²: 53% (Fig 1), the OR was: 2.00 (95% CI: 1.13 to 3.53) I²: 45% (Supplemental Fig 6) and RR: 1.21 (95% CI: 1.01 to 1.44) I2:64% (Supplemental Fig 7) in favor of the ultrasound-assisted group.

This definition was used in 2 studies^21,22  (69 participants) with conflicting results. The calculated RD was: 8.0% (95% CI: −35% to 51%) I²: 78% (Fig 1), the OR was: 1.88 (95% CI: 0.09 to 39.16) I²: 78%, (Supplemental Fig 4) and RR:1.1 (95% CI: 0.63 to 1.91) I²: 72% (Supplemental Fig 5).

This definition was used in 3 studies^23,25,27  (373 participants). The calculated RD was: 13% (95% CI: −1% to 28%) I²: 53% (Fig 1), the OR was: 1.73 (95% CI: 0.95 to 3.15) I²: 49% (Supplemental Fig 6), and RR: 1.29 (95% CI: 0.93 to 1.8) I²:60% (Supplemental Fig 7), in favor of the ultrasound-assisted group.

A different definition was used in the study by Ozdamar et al²⁴ : <400 RBC/μL. In the study by Kim et al,²⁶  authors did not provide the definition of successful LP.

We included 7 RCTs^21–27  with 618 participants, regardless of the definition of traumatic LP used by the authors in this analysis.

Overall, there were 35 traumatic LPs out of 295 total LPs (11.9%) in the ultrasound-assisted group and 64 traumatic LPs out of 282 total LPs (22.7%) in the standard group. The calculated RD was: −12% (95% CI: −22.0% to −0.3%) I²: 61% (Fig 2). The OR was: 0.45 (95% CI: 0.26 to 0.78), I²: 22% (Supplemental Fig 8) and RR: 0.53 (95% CI: 0.35 to 0.79), I²:8% (Supplemental Fig 9) in favor of the ultrasound-assisted group.

This definition was used in only 2 studies^21,22  (61 participants). The calculated RD was: −15% (95% CI: −30% to −1%) I²: 0% (Fig 2) OR was: 0.25 (95% CI: 0.04 to 1.61) I²: 0% (Supplemental Fig 6) and RR: 0.29 (95% CI: 0.05 to 1.61) (Supplemental Fig 7).

This definition was used in 3 studies^23,25,27  (341 participants). The calculated RD was: −10% (95% CI: −29% to 9%) I²: 81% (Fig 2) OR was: 0.58 (95% CI: 0.23 to 1.5) I²: 62% (Supplemental Fig 8) and the RR: 0.66 (95% CI: 0.33 to 1.3) I²: 53% (Supplemental Fig 9).

A different definition was used in the study by Ozdamar et al:²⁴  >400 RBC/microL, and in the study by Kim et al,²⁶  authors did not provide the definition of traumatic LP.

This outcome was reported in 4 RCTs^21,25–27  (212 participants). Calculated mean difference was: −1.11 minutes (95% CI: −2.88 to 0.66), I2: 77% showing that in both groups time for the LP procedure was similar (Supplemental Fig 10).

Failure of the LP defined as no CSF obtained during the procedure was assessed in 6 studies.^{21–23,25–27}  In the study by Ozdamar et al,²⁴  there were no failed LP.

Overall, there were 17 failed LPs out of 253 total LPs (6.7%) in the ultrasound-assisted group and 38 failed LPs out of 254 total LPs (14.9%) in the standard group. The calculated RD was: −7% (95% CI: −17.0% to −0.3%) I2: 72% (Supplemental Fig 11). The OR was: 0.43 (95% CI: 0.16 to 1.17), I2: 49% and RR: 0.46 (95% CI: 0.2 to 1.07) I2: 41% in favor of the ultrasound-assisted group.

This meta-analysis of 7 RCTs (618 participants) demonstrated that preprocedural ultrasound increased the success rate of first-attempt LP and reduced the incidence of traumatic LPs compared with standard procedure in children regardless of the definition of successful LP used. Additionally, the mean time of the LP procedure was similar in both groups. The heterogeneity for the main outcomes was slightly above the threshold of significance. The methodological quality of the included studies was moderate to low. The doubts most often concerned blinding of the participants, personnel, and outcome assessment.

In our analysis, we decided to combine included studies regardless of the definition of the successful LP use by the authors. Sample of CSF between 1000 RBC/μL and 10 000 RBC/μL can still be considered as a successful one. First of all, definition of <10 000 RBC/μL for the successful LP was used in the recent and largest study (1082 infants) evaluating the influence of the position of the patient during LP (sitting position versus lying) and it is still considered as up to date.²⁸  In the era of PCR testing, even CSF samples contaminated with RBC can be safely used to exclude meningitis without compromising diagnostic accuracy.²⁹  Additionally, such samples can be safely used to perform a CSF culture.¹  In contrast to the systematic review by Olowoyeye et al¹³  (4 RCTs - no effect in increasing LP success), by Gottlieb et al¹⁴  (6 RCTs - borderline significance in increasing LP success in the pediatric group), and in the most recent review by Kuitunen et al¹⁵  (7 RCTs - borderline significance in increasing LP overall success, improvement in first attempt LP success rate but only in infants), we present a novel approach of combining existing evidence on the effectiveness of the POCUS in first attempt success rate LP. A promptly obtained CSF sample in conjunction with the transformation in diagnostic methodologies, such as the utilization of PCR testing, enables the safe and quick exclusion of meningitis, even when using contaminated CSF samples.

We then decided to present results using RD, OR, and RR to enable the direct comparison of the results from previously published reviews. RD corresponds with absolute risk reductions, and it is more straightforward to interpret and provides more directly relevant information than relative measures. OR provides clinicians with a measure of association between exposure and outcomes but it may be difficult to interpret. RR is the most common statistic used in the RCTs and it was used in the most recent review by Kuitunen et al¹⁵  and by Olowoyeye et al¹³  (but in this review, authors focused only on the failure rate of the LP).

The use of ultrasound before the LP procedure gives the opportunity to determine the best intervertebral space possible to puncture. The puncture point may be different from the one determined by palpation because ultrasound allows assessment of where the spinal cord ends and gives more options to choose the best intervertebral space. Furthermore, ultrasound can help to identify any anatomic structures such us venous plexuses which should be avoided by the physician and helps to estimate the depth to which the needle needs to be inserted, which is very important for inexperienced people. Ultrasound-guided LP is accessible, easy to learn, does not bring any harms to the patients, and it does not prolong the LP procedure.³⁰  Among included studies, ultrasound assessment was done mostly by the physicians, reflecting the conditions under which this procedure of POCUS should take place.

In this systematic review, we performed a comprehensive literature search, with extended search strategy in major databases. We followed rigorous guidance from the Cochrane handbook and PRISMA statement. However, our review has some limitations. First, different definitions of successful and traumatic LPs were used in the included studies, mostly a threshold between 1000 RBC and 10 000 RBC were used. Those differences, together with the different ages of the participants, were largely responsible for the sources of heterogeneity. Second, more than a half of included studies had small sample size and, in most studies, a clinically relevant information, such as BMI of the patient, duration of antibiotic exposure and length of hospital stay, experience of the ultrasound provider, were not collected. Those would have brought crucial data to select a group of patients who especially benefit from pre procedural ultrasound. Among the included studies, 1 was published in in Korean that could be a source of bias, however most important data were published in English.²⁶  All of the studies were performed in pediatric emergency departments so the findings cannot be extrapolated to the neurology or intensive care wards patients.

Future trials should focus on recruiting higher numbers of participants with a preplanned protocol to collect more clinically relevant data. Additionally, improving randomization together with the application of a correctly conducted blinding procedure would provide high-quality evidence.

This systematic review and meta-analysis showed that POCUS improved the first attempt success rate and reduced the incidence of traumatic LPs compared with the standard LP procedure in children, regardless of the definition used. Because of the lack of harms, no impact on the time of procedure, and ease of implementation, if it is available, POCUS should be routinely used before every LP. To be decisively recommended, well‐conducted RCTs assessing its effectiveness should be performed.