OBJECTIVE. To assess the effects of procedural techniques, local anesthetic use, and postgraduate training level on lumbar puncture (LP) success rates.
METHODS. In this prospective observational study, medical students and residents (“trainees”) reported techniques used for infant LPs in an urban teaching emergency department. Data on postgraduate year, patient position, draping, total and trainee numbers of attempts, local anesthetic use, and timing of stylet removal were collected. Logistic regression analysis was used to identify predictors of successful LP, with success defined as the trainee obtaining cerebrospinal fluid with <1000 red blood cells per mm3.
RESULTS. We collected data on 428 (72%) of 594 infant LPs performed during the study period. Of 377 performed by trainees, 279 (74%) were successful. Local anesthesia was used for 280 (74%), and 225 (60%) were performed with early stylet removal. Controlling for the total number of attempts, LPs were 3 times more likely to be successful among infants >12 weeks of age than among younger infants (odds ratio [OR]: 3.1; 95% confidence interval [CI]: 1.2–8.5). Controlling for attempts and age, LPs performed with local anesthetic were twice as likely to be successful (OR: 2.2; 95% CI: 1.04–4.6). For infants ≤12 weeks of age, early stylet removal improved success rates (OR: 2.4; 95% CI: 1.1–5.2). Position, drape use, and year of training were not significant predictors of success.
CONCLUSIONS. Patient age, use of local anesthetic, and trainee stylet techniques were associated with LP success rates. This offers an additional rationale for pain control. Predictors identified in this study should be considered in the training of physicians, to maximize their success with this important procedure.
One of the first procedures learned by pediatric and emergency medicine residents is the lumbar puncture (LP). For febrile infants needing cerebrospinal fluid (CSF) evaluation, failure to obtain fluid or contamination with peripheral blood (“traumatic LP”) can result in unnecessary hospitalizations or a prolonged antibiotic course. Unsuccessful and traumatic LPs are especially common when procedures are performed by trainees, occurring 20% to 55% of the time.1–3
Although the importance of a good holder is legendary, other controllable factors influencing LP success have not been well evaluated. Use of topical anesthetics for an emergency department (ED) population improved attending physician success rates in one abstract,4 but other studies in teaching hospitals found that injected lidocaine did not help1 or even hurt3 success rates. To our knowledge, other techniques, including patient positioning (sitting or decubitus position)5 and drape removal, have not been evaluated with respect to LP success rates.
One technique purported to improve LP success rates involves removal of the stylet immediately after passage through the epidermal and subcutaneous tissues (early stylet removal, also called the Cincinnati method).6 The needle is then advanced through the dura without a stylet, theoretically allowing observation of CSF flow immediately after the subarachnoid space is entered. Although several case series allude to this method,7,8 to our knowledge there are no prospective studies addressing the prevalence of its use, its effects on success rates, or complications.
The purposes of this study were to identify variations in technique that were independently predictive of trainees' LP success. The primary study hypotheses were that early stylet removal and local anesthetic use would be associated independently with successful LPs among infants.
Study Sample Population
We conducted this prospective observational study from October 1, 2001, through February 25, 2003, in the ED of an urban pediatric teaching hospital (Children's Hospital of the King's Daughters, Norfolk, VA). The study was approved, with informed consent waived, by the institutional review board of Eastern Virginia Medical School.
Our 40000 patient visits per year are triaged to an attending physician-run fast track or a teaching ED. Almost all of the 25000 patients triaged to the ED are examined first by trainees, including fourth-year medical students and pediatric, emergency medicine, and family practice residents. Of trainee shifts, pediatric residents cover two thirds, emergency medicine residents cover one fourth, and family practice residents or fourth-year medical students cover the rest. Surveys administered previously at our institution indicated that pediatric and emergency medicine residents have performed >50 LPs by their third year of postgraduate training.
The typical LP protocol is for a resident to examine the patient, to order laboratory tests, and then to present the patient to an attending physician. For neonates with fever, blood and urine collection and any topical anesthetic applications are initiated before presentation and LP. Our hospital uses Cardinal Health pediatric/infant Safe-T-LP sets (Cardinal Health, Dublin, OH) with steel-hub, 1.5-inch, 22-gauge, spinal needles. ED technicians (paramedics) hold the infants for all LPs. At the time of the study, the holders had between 5 and >15 years of ED experience.
The patient study population included all infants from birth to 12 months of age who required LP as part of an ED evaluation. Exclusion criteria included prior LPs for the study, prior LPs at another hospital before transfer to our institution, congenital abnormalities of the lumbar spine, history of a bleeding disorder, and evidence of vasculitis or cellulitis over the lumbar spine. Children with subsequently diagnosed subarachnoid bleeding from a traumatic brain injury or herpes simplex infection were excluded from the analysis.
Local anesthetics available during the study period included lidocaine, a eutectic mixture of local anesthetics (EMLA LMX4; Ferndale Laboratories, Ferndale, MI), and a 4% liposomal formulation of lidocaine (AstraZeneca, Wilmington, DE). When ordered, topical local anesthetic is applied before blood and urine collection for patients undergoing evaluation for sepsis. If anesthetic is not in place long enough to be effective, then lidocaine use is encouraged but routine LP delay is not.
Questions regarding LP techniques were written on a 1-page form attached to all LP trays. To evaluate any influence of reporting bias on success and to assess the interrater reliability of reported techniques, 10% of trays had a duplicate form in another color. The duplicate forms required a fellow or attending physician to observe the entire procedure and to record LP data independently. Trainees were not directed to perform LPs with any particular technique and were blinded with respect to the hypotheses of the study. Medical students, residents, fellows, and attending physicians recorded anonymously their level of training, the patient's age and medical record number, the position of the infant, the use of anesthetics, whether the paper drape was over the patient, the method of LP (early stylet removal or traditional), and the numbers of needles used and attempts required before CSF was obtained. If the initial physician was unable to obtain CSF, then the number of subsequent attempts by an upper-level physician was recorded. Holder experience was not recorded. Completed forms were placed in a collection box in the ED.
The primary outcome measure was LP success, defined as the trainee obtaining adequate CSF for culture and cell count with <1000 red blood cells per mm3.1–4 When a trainee asked another resident, fellow, or attending physician to obtain CSF, the LP was considered a failure. If the CSF obtained was too traumatic to distinguish from peripheral blood or too scant to yield cell count data, then it was sent to the laboratory for “culture only,” which also counted as a failure. Data were obtained weekly from the hospital computer system to determine the total numbers of CSF cultures and cell counts from all ED infant LPs. In no case was CSF sent for other studies but not culture.
Factors predicting success were modeled with χ2 and multivariate logistic regression analyses with backward elimination. To determine the agreement between the practitioners' and observers' reported data for the same LPs, κ values were calculated with 95% confidence intervals (CIs) for nominal data and Spearman's correlation coefficients were calculated for ordinal data. The baseline success rate of the traditional LP technique was assumed to be 60% for an infant ED population.1,3 Detection of an absolute difference of 15% in success rates with 80% power required 152 patients in each group for any given predictor. Statistical analyses were performed with the SAS software package (SAS Institute, Cary, NC).
During the study period, 594 infant LPs from the ED yielded enough CSF for a culture or had data forms collected. Data forms for 428 infant LPs (72%) were collected. One patient was excluded because of previous enrollment, 4 because of LPs at other hospitals, 1 because of inflicted traumatic brain injury, and 1 because the medical record number and date of birth were not recorded. There were no diagnosed cases of herpes encephalitis in this population.
Overall success rates did not differ significantly between LPs with and without completed data forms (χ2 = 1.26) (Fig 1). Success rates during the first 3 months of the study (October to December 2001) were not statistically different from those during the last 3 months of the study (December 2002 to February 2003).
Twenty-nine LPs were performed by attending physicians or fellows and were not included in the remainder of this analysis. Of the remaining 392 LPs, 15 forms did not indicate training level. The total number of LPs available for analysis of trainee techniques was 377.
The overall success rate for LPs performed by medical students or residents was 74% (SE: 0.0226; 95% CI: 69.6–78.4%), and rates differed significantly according to the age of the patient (Table 1). Trainees reported using early stylet removal for 225 LPs (60%) and traditional stylet removal for 149 of 374 LPs; 3 forms were missing method data. Some type of local anesthetic was used for 280 (74%) of 377 LPs. In χ2 analyses, there was no difference in success rates according to trainee level, anesthetic use, early stylet removal, drape use, or infant position.
In addition to age, the total number of attempts by all providers was identified as a strong confounder, with goodness-of-fit testing indicating that modeling this factor as a continuous variable resulted in the best model fit (χ2 likelihood ratio: 192; P < .0001). Multivariate logistic regression analysis with backward elimination was used to evaluate the independent contribution of each factor to the success of the LP. Variables examined included total anesthetic use, stylet technique, position, drape, training level, and patient age, with the cutoff point for age of >12 weeks or <12 weeks maximizing model fit. Results of the logistic regression model after elimination of the least-significant variables, with α to remain set at P = .1, are presented in Table 2. For the subset of patients <12 weeks of age, early stylet removal improved success rates significantly (odds ratio [OR]: 2.4; 95% CI: 1.1–5.2).
Observers were present for 24 (6.4%) of 377 LPs and simultaneously recorded the same data as the person performing the LP. One of the 24 LPs was missing method information from either the practitioner's or observer's data record. With regard to LP stylet method, 21 of 23 agreed (κ = 0.74; 95% CI: 0.41–1.08). The κ values for all other nominal variables (LP position, topical anesthetic, and drape use) ranged between 0.70 and 1, except for lidocaine use (κ = 0.62). The weighted κ values were 1.0 for both the number of attempts and the number of needles used. The success rate for observed LPs was 75%, which was not significantly different from the overall success rate.
In the past 5 decades, most pediatric LP literature has dealt with the implications of CSF results, rather than the procedure itself. Articles that discuss procedural techniques have done so in the context of hospitalized neonates,1,2,5,9,10 rather than febrile infants in the ED.3,4
For such a common procedure, there is a surprising lack of controlled data underlying textbook recommendations for procedural nuances such as bevel orientation, patient position, and pain control. Only the need for a stylet at insertion has fairly robust support.7,8
Findings in the infant pain control literature have been contradictory.1,3,11 Although pediatric procedure textbooks dictate that “withholding local analgesia for this procedure is strongly discouraged,”12 establishing the efficacy of pain relief for neonatal LPs has been problematic. Kaur et al10 performed the only study that demonstrated diminished heart rate elevation and facial pain response, randomly assigning 60 neonates to topical local anesthetic (EMLA) or placebo. Other studies used similar physiologic data and failed to show a statistical difference, perhaps because the pain response was overwhelmed by the effect of positioning.1,9 Lack of demonstrated or perceived pain control may be why >50% of pediatric emergency medicine physicians surveyed do not use analgesia for patients <3 months of age.13 What has not been established prospectively is whether local anesthesia influences LP success.
Pinheiro et al1 found injected lidocaine to be neutral for success, although it decreased neonatal struggling. Among 99 children 0 to 36 months of age, Carraccio et al3 found that injected lidocaine actually increased the rate of traumatic LPs (1000 cells per high-power field). If this were attributable to loss of palpable bony landmarks, then one would not expect a topical anesthetic to cause the same problem. Indeed, in the topical anesthetic study by Kaur et al,10 2 of 30 LPs were traumatic in the EMLA group and 3 in the placebo group. One abstract by Shenkman et al4 found EMLA to reduce the number of traumatic LPs by ED attending physicians, compared with a historical cohort, for patients <2 months of age (OR: 0.31; 95% CI: 0.13–0.73). Ours is the first prospective study that found the use of pain control to be predictive of increased success rates and the only study that examined resident LP performance explicitly. Because only 4% of anesthetic users injected lidocaine in our study, the potential confounding issue of loss of landmarks was minimal.
Beyond describing the relationship between pain control and infant LPs, we were interested in quantifying and describing the technique of early stylet removal. Historically, adult LPs were performed with a stylet in place, to avoid spinal headache and obstruction of the needle with tissue. By the middle 1950s, it had become common practice among children to omit the stylet to improve success rates.8 In 1956, Choremis et al14 connected 5 cases of pediatric spinal epidermoid tumors to repeated use of nonstyletted LP needles in tuberculous meningitis. Multiple case reports supporting the association between even a single LP and the development of these tumors 1.5 to 23 years later were published.7,8,14–16 This ultimately led to an editorial practice guideline stating that “a stylet should always be used”17 and censure15 of Greensher et al,18 who stated that the use of butterfly needles gave their group a 100% success rate without complications over 2.5 years of observation. Partially in response to observations by Greensher et al,18 Schreiner and Kleiman2 conducted a prospective study of 246 LPs among 181 NICU patients and found no difference in the rates of traumatic LPs with styletted spinal needle versus butterfly needle usage.
Although current texts recommend the traditional LP method with stylet always in place,19,20 the concept of early stylet removal once past the epidermis is not discussed. This method was mentioned as a “perfectly safe” option in the epidermoid tumor case reports.7,8 In 1992, Bonadio6 published a review referring to early stylet removal as the “Cincinnati method,” which was purported to reduce the risk of unsuccessful LPs, but cited no data. In a recent national survey of pediatric emergency medicine physicians, 16% of 188 stated that they prefer early stylet removal and almost twice that number sometimes advance the needle without a stylet in place.13
Because early stylet removal has never been formally studied, it is difficult to know whether it is safe or effective. Gibson and Norris21 found that ∼70% of needle insertions through skin without a stylet yielded a core of epidermis, but they noted that there were only rarely fat fragments. Because subcutaneous tissues are endodermal, not ectodermal, and because fat is “cored” infrequently by beveled needles, early stylet removal should carry little risk of tumor formation. A difficulty is in knowing exactly when the practitioner is past the epidermis.
Our findings indicated that, for younger infants, early stylet removal was associated with increased success rates. Logically, the method's advantages should be more pronounced in situations in which the dural “pop” is subtler and the diameter of the spinal canal is smaller. Assessing safety would require long-term follow-up monitoring for a very rare event and would be particularly difficult because use of this method might not be documented explicitly.
Our study was limited by several factors. Most notably, because we did not dictate who used which method, we could perform neither randomization nor control. Thought was given to having residents perform procedures with and without various techniques, to serve as their own control subjects, but residents indicated a strong belief in the merits of their chosen preferences. In particular, many considered withholding anesthetic to be unethical. We decided that success with forced unfamiliar techniques might bias performance more than letting all practitioners perform their perceived best techniques.
Another limitation could be reporting bias based on our outcome measure, ie, success. Success rates for LPs with completed forms and without forms were not significantly different, however, even assuming proportional numbers of “dry” LPs without laboratory sample collection. Our attempt to assess reporting bias by using interrater assessment fell short, with only 6.4% of LPs being observed, rather than the 10% of kits with duplicate forms. When an attending physician was unavailable to observe the procedure in a timely manner, many residents performed the LP anyway or chose a kit without double forms. Because the residents could not know their success or failure before the observed LP performance, this should not have biased reporting. Rather than revealing decreased success rates, observed LPs were slightly more likely to be successful. This could be attributable to coaching by the upper-level observer, because their instructions did not prohibit them from giving advice.
We did not collect data on the infant holder or on characteristics of individual practitioners beyond the year of training. An experienced holder might be able to coach an inexperienced resident to success or an inexperienced holder might allow more patient movement and decrease success rates. However, because the holders were assigned after the topical anesthetic decision was made, these results should not have been influenced. Type of residency and number of previously performed LPs could confound the effects of pain and stylet manipulation. The lack of statistical difference in success rates according to postgraduate year might simply indicate a plateau in the learning curve for all practitioners.
The majority of studies on which we based our power analysis involved neonates. In light of our finding that success rates improved with infant age, the inclusion of older infants decreased the power to detect a difference. Given our large number of successful “cases” versus controls, the logistic regression model amply supported the number of variables modeled.22 However, to establish prospectively the efficacy of early stylet removal among older infants, a much larger number would be needed.
The confounding relationship between the total number of attempts and decreased success rates could simply reflect more blood in the CSF after multiple attempts. We propose that the number of attempts may be a proxy variable for the inherent difficulty of LP. This is supported by the fact that the number of attempts was modeled best as a continuous variable, rather than a categorical variable.
Both pain control and early stylet removal decreased the rates of failed LPs, controlling for number of attempts and age. Although we cannot recommend primarily teaching the early stylet removal method, because of lack of safety data, instructors training residents should be aware of this technique, particularly for young infants. Our data indicate that local anesthesia is feasible for a large number of febrile infants who require LP. In addition to compassionate considerations, the association with improved success offers another rationale for the use of pain control.
- Accepted July 19, 2005.
- Address correspondence to Amy L. Baxter, MD, Pediatric Emergency Medicine Associates, PO Box 422002, Atlanta, GA 30342. E-mail:
This research was presented in part at the American Academy of Pediatrics meeting; October 19–23, 2002; Boston, MA.
Dr Isaacman's current affiliations are: Department of Pediatrics, University of Pennsylvania School of Medicine, Philadelphia, PA; and Clinical Affairs, Vaccines, Global Medical Affairs, Wyeth Pharmaceuticals, Collegeville, PA 19426.
Financial Disclosure: Dr Baxter has received donated EMLA from AstraZeneca. There are no other financial relationships relevant to this article to disclose.
- ↵Pinheiro JM, Furdon S, Ochoa LF. Role of local anesthesia during lumbar puncture in neonates. Pediatrics.1993;91 :379– 382
- ↵Shenkman A, Jukuda J, Benincasa G, et al. Incidence of traumatic lumbar puncture in children treated with EMLA at a pediatric emergency room [abstract]. Pediatr Emerg Care.2002;18 :395A
- ↵Gleason C, Martin R, Anderson J, Carlo W, Sanniti K, Fanaroff A. Optimal position for a spinal tap in preterm infants. Pediatrics.1983;71 :31– 35
- ↵Bonadio WA. Interpreting the traumatic lumbar puncture. Contemp Pediatr Res Q.1992;1 :23– 32
- ↵Halcrow SJ, Crawford PJ, Craft AW. Epidermoid spinal cord tumor after lumbar puncture. Arch Dis Child.1985;60 :978– 979
- ↵Porter FL, Miller JP, Cole FS, et al. A controlled clinical trial of local anesthesia for lumbar punctures in newborns. Pediatrics.1991;88 :663– 669
- ↵Enad D, Salvador A, Brodsky NL, et al. Safety and efficacy of eutectic mixture of local anesthetics (EMLA) for lumbar puncture in newborns [abstract]. Pediatr Res.1995;37 :204A
- ↵Cronan KM, Wiley JF II. Lumbar puncture. In: Henretig FM, King C, eds. Textbook of Pediatric Emergency Procedures Baltimore, MD: Williams & Wilkins; 1997:546–547
- ↵Choremis C, Economos D, Papadatos C, et al. Intraspinal epidermoid tumours (cholesteatomas) in patients treated for tuberculous meningitis. Lancet.1956;2 :437– 439
- ↵Carlson DW, Digiulio GA, Gewitz MH, et al. Illustrated techniques of pediatric emergency medicine. In: Fleischer G, Ludwig S, eds. Textbook of Pediatric Emergency Medicine 4th ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2000:1813
- ↵Barkin RM. Meningitis, bacterial. In: Barkin RM, ed. Pediatric Emergency Medicine Concepts and Clinical Practice St Louis, MO: Mosby-Year Book; 1992:915
- Copyright © 2006 by the American Academy of Pediatrics