Pre-procedural Ultrasound for Infant Lumbar Puncture: a randomized clinical trial

Abstract

Background

Our purpose was to determine the potential effect of pre-procedural ultrasound to increase lumbar puncture (LP) success compared with standard palpation method. Further, we assessed feasibility of and clinician satisfaction with a standardized ultrasound protocol.

Methods

This prospective, two-arm, parallel group randomized trial was conducted in a single-center pediatric emergency department. We compared pre-procedural ultrasound vs. palpation method on success with infant LPs. Infants <3 months of age requiring LP were enrolled. Sixteen pediatric emergency medicine physicians with varied ultrasound experience were trained to conduct the ultrasounds to mark interspace locations. Primary outcome was successful LP, defined as obtaining a cerebrospinal fluid (CSF) sample on first attempt with <1,000 red blood cells per high-powered field (clear CSF). Secondary outcomes included clear CSF on any attempt, any CSF on the first attempt, traumatic LP proportion, and LP attempt frequency. Feasibility was assessed by comparing provider number attempting the LP and procedure duration. Clinician satisfaction and sonographer perceptions of ultrasound acceptability and impact were assessed.

Results

Eighty-one patients consented and 80 were analyzed (99%): 40 per group. No statistical difference was seen for the primary outcome (p >0.05) between intervention and control groups (difference 3%; 95% CI -19% to 24%). There were no statistical differences between intervention and controls groups for secondary outcomes including the rate of traumatic LPs, number of attempts, and the duration of LP procedure. Most sonographers (84%) strongly agreed or agreed that the US protocol technically easy to perform, well tolerated by the patient (94%), well accepted by the family (100%), and well accepted by the LP procedural clinicians (99%). In the US group, the majority of clinicians who performed the LPs (68.4%) noted that the pre-procedural US influenced their behavior, most commonly helping with overall visualization at the selected interspace (28.9%) or prompting a change in interspace (26.3% higher, 5.3% lower). Seventy seven percent agreed or strongly agreed that they would like to use the technique again for their next LP. The mean ultrasound duration was 4.6 minutes.

Conclusions

Pre-procedural US by did not improve the rates of first attempt success when compared with palpation method. Our results suggest ultrasound is feasible and well accepted, with a perceptible impact on care.

Introduction

Fever is among the most common reasons young infants present to emergency departments (EDs).^1,2 In febrile infants younger than 60 days, lumbar puncture (LP) is a procedure commonly performed to evaluate for meningitis. However, success with the procedure is often sub-optimal.³ Using the standard anatomic palpation technique (palpation method), providers are frequently unable to obtain diagnostically useful cerebrospinal fluid (CSF) on the first infant LP attempt. Novice operators are unsuccessful in over 60% of LPs.³ Even experienced providers fail to obtain clear fluid on first attempt 37% of the time.⁴

There are several potential adverse effects from each unsuccessful LP attempt: prolonged pain, extended restraint, and the potential for bloody, or traumatic LPs. Traumatic LPs can occur in up to 50% of LP attempts.^4,5 Traumatic LPs are difficult to interpret, leading to diagnostic uncertainty, prolonged hospital stays, and potential iatrogenic complications.^6–11 Reductions of unsuccessful and traumatic LPs in infants can improve diagnostic ability and reduce patient harm.

Ultrasound (US) has the potential to increase LP success rates and decrease traumatic LPs but has not been adequately studied in pediatric patients. US can also help visualize the termination of the conus (which can range from the T12 to the L4 interspace in infants) in order to safely avoid the spinal cord.^12–18 Two prior randomized trials showed promising results for US-assistance compared to standard techniques for interspace selection. However, these studies were single-centered and used few sonologists, and could therefore benefit from further studies to confirm their reproducibility and generalizability.^19,20

In the current pilot study, we aimed to determine the potential effect of performing pre-procedural US to increase LP success rates when compared with the palpation method. We were interested in understanding the potential effect of pre-procedural US under real-life circumstances, when any of a number of clinicians would be performing the US and conducting the LP procedure. Further, we aimed to determine feasibility of and clinician satisfaction with a standardized US scanning protocol for pre-procedural infant LP.

Methods

Study design

We conducted a single-center, prospective, two-arm, parallel group randomized clinical trial. Ethical approval was obtained from the institutional review board with written informed consent obtained from the legal guardians of all participants. The study was registered with clinicaltrials.gov (identifier # NCT02373774).

Study setting and population

We conducted this trial in an urban pediatric emergency department with an annual volume of approximately 50,000 patients. Infants younger than 90 days were included if their diagnostic workup included a LP and a sonologist was available. We excluded patients if they had a prior LP, were clinically unstable, or had a known history of spinal dysraphism.

Study protocol

We block-randomized participants 1:1 in permuted blocks of 4 and 6 into either the 1) visualization with pre-procedural lumbar US (experimental) group or the 2) palpation method (control) group. We concealed allocation using a password protected website with allocation conducted online at the time of enrollment.

In the experimental group, the clinician who would perform the LP (termed the “LP procedural clinician”) first selected a lumbar interspace using the palpation method. Subsequently, the study sonologist conducted the US protocol (supplementary figure 1a), which consisted of all of the following: 1) an assessment for CSF at the spinal level selected by the LP procedural clinician (short axis); 2) an assessment of the number of interspaces above the chosen space at which CSF was found without the conus being present (short axis view); 3) a measurement of the depth at which CSF was encountered (long-axis view); and 4) an evaluation of any overlying vasculature (short axis view). The sonologist reviewed this information with the LP procedural clinician who made the final selection of which interspace to use for the LP. The selected interspace and location of the spinal cord was then marked by the sonologist using a sterile marking pen (supplementary figure 1b).

The standard of care (control) group used the palpation method to select a lumbar interspace for the LP procedure. Both groups had access to cognitive aids detailing the palpation method, which consisted of a one page illustrated guide demonstrating the landmarks used to help identify the L5 vertebrae (supplementary figure 2).

In order to include sonologists across a range of expertise, 16 pediatric emergency medicine attendings and fellows were trained on the experimental US scanning protocol with a one-hour didactic and 20-minute hands-on session. For both groups, the LP procedural clinician could be a pediatric resident, emergency medicine resident, pediatric emergency medicine fellow or a pediatric emergency medicine attending.

Research assistants measured the duration of the LP procedure for both study groups, defined as the time “from when needle breaks skin to when first fluid drop comes out OR procedure is stopped (including changing providers, repeating ultrasound, or abandoning procedure.)” In the experimental group, the duration of the US was also timed and defined as the time “from when probe is placed onto skin to when sonologist declares s/he is done”.

The LP procedural clinicians in both groups completed a questionnaire including technical details of the procedure (e.g. number of attempts taken, types of analgesia used, etc). In the experimental group, LP procedural clinicians also completed a questionnaire on their attitudes towards the pre-procedural US. Similarly, study sonologists completed a questionnaire to determine their attitudes towards how useful the US was as well as the perceived acceptability of the US to the team. The study research coordinator obtained laboratory data from the electronic health record.

Measures

Our primary outcome was successful LP, defined as obtaining a sample of CSF on the first attempt with a red blood cell (RBC) count of <1,000 per high-powered field (HPF).

Secondary outcomes included: 1) success obtaining CSF with <1,000 RBC/HPF (regardless of the number of attempts); 2) success obtaining any CSF (regardless of RBC count) on the first attempt; 3) the proportion of traumatic LPs, defined as CSF with greater than 1,000 RBCs per HPF on any attempt; and 4) the number of attempts until CSF was obtained or the procedure was abandoned, defined as the number of times a LP needle was removed from skin and reinserted or a new needle was inserted.

We assessed feasibility of the US protocol by: 1) measuring the duration of the US in minutes and 2) comparing the duration of the LP between groups, defined as the number of minutes between spinal needle first breaking skin and return of CSF in the spinal needle on all attempts (or abandonment of the procedure).

Satisfaction of the LP procedural clinician with the US protocol and preferences were assessed with Likert-style questions inquiring about the perceived impact of the US protocol. (e-appendix 1) Sonologist perceptions of US acceptability and impact on the procedure were assessed with Likert-style questions.

We collected potential confounders and modifiers of LP success including 1) provider variables (age, experience with US, experience with LPs, training level, 2) patient variables (age) and 3) procedural variables (early stylet removal, analgesia use, holder, parental presence, supervisor level, and patient position). Early stylet removal refers to the technique where the stylet is not reinserted after the skin is initially punctured until CSF is collected and the procedure is completed.

Sample size and Data Analysis

Based on funding, we enrolled for 12 months, with an expected enrollment of 100 patients (120 eligible and at least 80% consent rate). With a baseline LP success rate of 50% (from our prior work), 100 patients allowed us to detect a 28% or greater absolute increase in LP success with 80% power (p=.05, two sided).

We compared the primary outcome (success) and binary, categorical secondary outcomes between groups using absolute risk differences with 95% confidence intervals and the Pearson’s chi-squared test. For continuous secondary outcomes, we used the independent samples t-tests to compare means or a non-parametric equivalent for non-normal distributions. An intention-to-treat analysis was performed, with data imputed for any missing values.²¹ For outcomes analyzed as proportions, the observed proportion of all other participants was rounded to 0 or 1 and then imputed. For categorical outcomes, the category with the highest frequency, i.e. the largest proportion seen empirically, was imputed for that participant. These methods were applied to the one participant who had no LP performed after being allocated to the ultrasound group. All analyses were conducted using SAS software, Version 9.4.

Results

One-hundred and eleven eligible patients were approached and 81 were consented and randomized: 41 in the US group and 40 in the palpation group (Figure 1). A total of 63 unique LP procedural clinicians performed the 81 LPs. Fifteen sonographers (ranging in experience from 50 to 3000 lifetime sonograms performed) conducted a median of 2 LP sonograms for the study (interquartile range (IQR) 1 to 3). One sonographer did not participate.

Figure 1.

Study Flow Diagram

Table 1 compares the group characteristics. Training level and experience of the LP procedural clinicians with both infant LPs and US were similar between groups. Factors related to the LP procedure itself were also similar between groups, with the exception of slight differences in training levels of the LP supervisors, and greater use of analgesia and the early stylet removal technique in the control group (no statistically significant differences).

Table 1.

Group Characteristics

	Ultrasound group n=39	Palpation group n=40
Patient characteristics
Age in days, mean (SD)	29.9 (19.3)	25.4 (19.4)
LP procedural clinician characteristics*
Number of infant (<90 days) LPs performed, median (IQR)	5.0 (3.0, 12.0)	3.0 (2.0, 11.0)
Number of LPs performed in any age group, median (IQR)	8.0 (5.0, 13.0)	5 (3.0, 14.0)
Any prior experience performing spinal ultrasound for infant LP n (%)	3.0 (7.7)	9.0 (22.5)
Training level of procedural clinician, n (%)
Resident+	30 (76.9)	31 (77.5)
PEM Fellow	5 (12.8)	5 (12.5)
Attending	4 (10.3)	4 (10.0)
LP procedure factors
Holder for the LP procedure, n (%)
Technician	35 (89.7)	38 (95.0)
Nurse	3 (7.7)	2 (5.0)
Other	1 (2.6)	0 (0)
Analgesia use++
None	3 (7.7)	1 (2.5)
Oral	29 (74.4)	37 (92.5)
Topical	35 (89.7)	35 (87.5)
Local	1 (2.6)	5 (12.5)
Training level of supervisor, n (%)
No supervisor	8 (18.4)	11 (27.5)
Attending	27 (69.2)	21 (52.5)
PEM Fellow	4 (10.3)	8 (20.0)
Family members present during the procedure, n (%)	21 (53.8)	23 (57.5)
Used early stylet removal technique, n (%)	31 (79.4)	36 (90.0)

^*LP procedural clinician = person who performed the first attempt at infant LP

⁺Resident = post-graduate year 1, 2, or 3

⁺⁺Oral = oral sucrose, Topical = 4% lidocaine (LMX), Local = 2% lidocaine subcutaneous infiltration

SD= standard deviation, IQR=interquartile range, PEM = Pediatric Emergency Medicine, LP= Lumbar puncture

Outcomes

Table 2 compares outcomes between the groups. For our primary outcome of obtaining CSF with <1,000 RBC/HPF on the first attempt, improvement with pre-procedural US group was small, non-significant, with wide 95% CIs. Larger potential effect sizes favoring the intervention group were seen for obtaining CSF with <1,000 RBC/HPF (regardless of the number of attempts) and success obtaining any CSF on the first attempt (regardless of RBC count). For all outcomes, results remained the same when adjusting for potential confounding variables and assessing for any possible modifiers.

Table 2.

Efficacy Outcomes

	Ultrasound group n=40	Palpation group n=40	Absolute risk difference (95% confidence interval)	P-value*
Primary outcome
CSF obtained on 1st attempt with <1,000 RBCs, n (%)	19 (47.5)	18 (45.0)	3% (−19% to 24%)	0.82
Secondary outcomes
CSF obtained on any attempt with <1,000 RBCs, n (%)	30 (75.0)	26 (65.0)	10% (−10% to 30%)	0.33
CSF obtained on 1st attempt, n (%)	22 (55.0)	18 (45.0)	10% (−12% to 32%)	0.37
CSF obtained on any attempt, n (%)	37 (92.5)	37 (92.5)	0% (−12% to 12 %)	1.00
Number of attempts to obtain CSF or procedure abandoned, Median, IQR	2 (1-2)	2 (1-3)	—	0.25
The number of providers who attempted the LP, n (%):
One	29 (73.7)	23 (57.5)		0.37
Two	7 (15.7)	11 (27.5)
Three	4 (10.5)	6 (15.0)

^*two-sided test, significance for primary outcome p=0.05, significance for secondary outcomes included Bonferroni correction p=0.01

RBC = red blood cells, CSF = Cerebrospinal fluid, LP = lumbar puncture

Overall, there were fewer traumatic LPs (>1,000 RBC/HPF) in the intervention group (19.4% vs. 29.7% in the control group (95% CI for difference: -11 to 28%). Although the overall median number of attempts needed to obtain fluid was similar between groups, only 21% in the intervention group needed three or more attempts compared to 40% in the control group.

Feasibility and satisfaction

The US was completed in a median of 4.6 minutes (IQR 3 to 6.8). The duration of the infant LP was similar between groups, with a median of 1.6 minutes in the intervention group (IQR 0.8 to 13.4) and a median of 4.2 minutes for the control group (IQR 0.8 to 5.2).

Supplementary Table 1 shows the perceived impact, satisfaction, and acceptability of the US protocol. Most sonographers found the US protocol technically easy to perform, well tolerated by the patient, and well accepted by the family and LP procedural clinicians. In the US group, the majority of clinicians who performed the LPs (68.4%) noted that the pre-procedural US influenced their behavior, most commonly helping with overall visualization at the selected interspace (28.9%) or prompting a change in interspace (26.3% higher, 5.3% lower).

Discussion

In this pilot randomized controlled trial, no statistical difference was seen in first attempt acquisition of CSF with <1,000 RBC/HPF with use of pre-procedural US to visualize infant LP interspaces. Less stringent definitions of LP success (e.g. overall acquisition of CSF) also did not yield statistical differences. The US protocol was considered easy to perform, commonly influenced LP procedural approach, and did not increase the overall duration of the LP procedure.

The lack of clearly meaningful improvements in successful LPs in our study is in contrast with results from prior studies of similar size. In a single-center clinical trial (n=128), the investigators noted a much larger improvement in LP success (27%) when success was defined as CSF obtained on the first attempt with <1,000 RBC per HPF.²⁰ In that study, however, only two sonologists conducted the ultrasounds and the baseline LP success rates were notably lower (31%) than at our site. In the only other randomized trial of infant LPs, pre-procedural US led to a 3% decrease in LP success when success was defined as CSF on first attempt with <1,000 RBC per HPF (Michael Gorn, personal communication, 12/6/16). However, in that study, the investigator’s predefined primary outcome was CSF with <10,000 RBC on any attempt, with US use associated with an improvement in LP success of 27% and only three sonologists trained in the protocol.¹⁹

We chose a primary outcome that was particularly patient-oriented; however, collection of enough CSF for a culture (regardless of attempts or amount of RBCS) is certainly of clinical importance. In the two prior trials, pre-procedural US resulted in 9% and 18% increases in obtaining CSF on any attempts compared with only 2% difference in our study. ^19,20 The smaller difference we noted in this and other outcomes likely stems from our higher baseline rates of LP success using the palpation method and the variation in experience of the sonologists and those performing the LPs. The varying results from these three trials suggests improvement in LP success using pre-procedural US may depend on site specific characteristics, expertise, and the exact nature of the US procedure (which differed across studies). Our higher baseline rates of LP success with the palpation method also suggest the need to understand whether the benefit of US may be most realized in difficult cases. Our US protocol was most similar to that used by Neal et al. In their study, however, the investigators marked the skin with two intersecting lines directly over the interspace whereas we used a single perpendicular line alongside the selected interspace. This previous study also confirmed the interspace by counting vertebral bodies. In the study by Gorn et al. they similarly marked the interspace with a cross, and additionally indicated the measured ideal depth of insertion by marking the actual spinal needle with a sterile marking pen. These differences may have accounted for the differing results found in our study. ^19,20 Given the current body of literature and because the majority (77%) of procedural clinicians expressed that they would like to use ultrasound prior to their next infant LP, additional multi-center studies are warranted to establish clinical utility, appropriate protocols, and training.

Our results should help alleviate the concern that adding US to the overall LP procedure will increase the overall duration of the LP procedure. Prior studies have similarly noted that the US procedure took minimal extra time.^19,20 Another potential concern to use of US was that this new technique would not be well-accepted; however, we found the US to be well tolerated by patients, family members and LP procedural clinicians. Furthermore, procedural decision-making was regularly impacted by the US, leading to frequent changes in interspace selection. Interestingly, in a prior study, the investigators noted no meaningful differences in hospital length of stay and duration of antibiotic use between pre-procedural and palpation method groups, despite an increase in LP success using US.²⁰

Limitations

Although we standardized the US protocol and feedback to be provided to the LP procedural clinicians, we did not determine which specific components of the US were most influential on decision making with the LP procedure. Given the wide variability in sonologists, it is possible that they differentially emphasized specific US findings and thus had varying impact on decision making. Variable influence may have been further compounded by the fact that both LP procedural clinicians and sonographers were not blind to the intervention. Additionally, since this study took place in an academic setting, the majority of LPs were performed by trainees with little experience in conducting LPs and even less in utilizing US to assist the procedure (with more US LP experience in the control group). However, the high baseline success rate in the control group compared to the literature limited our ability to discern larger effect sizes. Unlike prior protocols, our sonologists did not choose which interspace to enter; rather the interspace was only marked once the decision was made by the procedural clinician regarding which interspace they wanted to attempt (given feedback from the sonologist). Finally, our small sample size limited our ability to detect small differences in the primary outcome and to determine any confounding or interaction effects. For example, there seemed to be slightly more frequent use of early stylet removal in the control group (90% vs 78.9%), which in past studies has been associated with improved success amongst novices.

Conclusions

Pre-procedural US did not lead to improved success on first LP attempt compared with the palpation method when performed by multiple sonologists and procedural clinicians. Our results do strongly suggest that pre-procedural US is feasible and well accepted, with a perceived impact on care.