There is substantial evidence in the medical literature that lumbar puncture (LP) is extremely safe [
There are only a small number of reports, however, suggesting catastrophic deterioration related to LP [
We designed this study to assess the prevalence of “high-risk” CT findings in patients who undergo head CT for any reason and in particular whether and how often such findings occur in patients whose clinical presentations make it extremely unlikely that they would be harmed by LP. Because of the absence of any gold standard for “high-risk” CT findings, we convened an expert panel and asked its member to define what CT findings indicate increased risk of tonsillar herniation following LP; we then used clinical data that had been prospectively recorded on a large cohort of patients for whom CT had been ordered to assess the relationship between selected history and physical examination (H&P) findings and the presence of those CT characteristics defined by the expert panel as conferring increased risk.
This study was approved by the University of California, Los Angeles (UCLA) Institutional Review Committee. Informed consent was not required as the clinical data collected was deidentified and did not impact clinical care.
We designed this study in two parts. First, given the absence of any prior criterion standard regarding which if any head CT findings predict increased hazard from performance of LP, we convened an expert panel to determine if it was possible for them to agree on any set of head CT characteristics that could be defined as predicting “high risk” of LP-induced brain hemiation (Part I). Second, we used prospectively collected data from the large NEXUS 2 cohort of emergency department patients for whom head CT had been ordered, for whatever reason, to evaluate how closely clinical findings correlate with the presence or absence of any of these “high-risk” CT findings (Part II). We did not make any assumptions regarding what would have happened to any of the
A structured e-mail survey regarding potential LP contraindications was sent to the panelists, each of whom then indicated his or her level of agreement (0 = “strongly disagree,” and 5 = “strongly agree”) with the following statement: “This CT finding should contraindicate performance of an LP,” for a series of possible CT findings (see Appendix information at “Information and Survey Sent to the Delphi Panel”). Narrative explanations of the experts’ responses were solicited, and panelists were asked to suggest any additional CT findings they felt should contraindicate LP. For subsequent rounds, survey participants were asked to read the anonymous narrative explanations submitted by all of the panelists and then to rerate each CT finding. While three survey rounds are generally considered adequate to achieve a valid consensus opinion [
Head CT abnormalities identified as contraindications to LP by the Delphi panel.
Head CT abnormalities | Percentage of the Delphi panel who rated the abnormality as 4 or 5 (out of 5) |
---|---|
Lateral shift of midline structures | 73 |
Loss of basilar cisterns | 73 |
Obliteration of the fourth ventricle | 91 |
Subtentorial herniation or tonsillar herniation | 100 |
Obliteration of the superior cerebellar cistern and the quadrigeminal plate cistern with sparing of the ambient cisterns | 36 |
We calculated the sensitivity, specificity, and positive and negative predictive values for each of the individual H&P findings in predicting the presence of increased risk of herniation with LP on head CT (as defined by the expert panel). We also analyzed the test characteristics of the complete neurologic examination and all aspects of the H&P, assuming that the presence of any single abnormal finding predicts increased risk as determined by CT.
The expert panel ultimately agreed on five findings on head CT that were felt to increase risk of LP sufficiently so as to contraindicate performance of this procedure (Table
Of the 1737 patients initially included, 445 had abnormal head CTs. Of these, 122 were excluded because of a previous neurosurgical intervention, leaving 323 patients for the final analysis. The average age of study subjects was 57.8 years (range, 1–99 years), and 58% were male. Other demographic characteristics are listed in Table
Demographic and clinical characteristics of study subjects,
Sex | |
Male | 188 (58.2) |
Age | |
Range | 1 to 99 |
Mean | 57.8 |
Median | 62 |
Interquartile range (IQR) | 47–87 |
Pediatric ages | |
≤3 | 8 |
3 to 6 | 3 |
6 to 18 | 11 |
Race | |
White | 189 |
Black | 17 |
Hispanic | 57 |
Asian | 22 |
Middle Eastern | 17 |
Others/unknown | 21 |
In categorizing the CT scan reports, both of the physician reviewers independently agreed that the written radiology report definitively indicated the presence, or absence, of the high-risk findings, in 68% of the cases. The emergency radiologist who then reviewed the actual CT images in the remaining 32% of cases was able to categorize findings as clearly present or absent in all but 2%; these few cases were then categorized by a senior emergency physician, using the conservative criteria described above (with cases classified as high risk only if they showed definitive dilation, enlargement or compression of the 4th ventricle or the brainstem, hydrocephalus, effaced cerebral sulci, local mass effect, or evidence of edema). Overall, 47 (14.6%) of the CT scans had at least one of the high-risk findings defined by the expert panel (Figure
Study subject selection.
None of the individual elements of H&P were sensitive in identifying patients with increased risk based on head CT findings (Table
Test characteristics of history and physical examination findings in relation to clinically abnormal head CTs.
|
Sensitivity | CI | Specificity | CI | NPV | CI | PPV | CI | |
---|---|---|---|---|---|---|---|---|---|
Blunt head injury | 139/307 | 47.7 | (32.5, 63.3) | 55.1 | (48.9, 61.3) | 86.3 | (80.2, 91.1) | 15.1 | (9.6, 22.5) |
Dangerous mechanism | 105/285 | 33.3 | (19.6, 49.6) | 62.6 | (56.1, 68.7) | 84.4 | (78.3, 89.4) | 13.3 | (7.5, 21.4) |
Vomiting | 23/318 | 8.9 | (2.5, 21.2) | 93 | (89.4, 95.8) | 86.1 | (81.6, 89.3) | 17.4 | (5.0, 38.7) |
Coagulopathy | 36/280 | 8.3 | (1.8, 22.5) | 86.5 | (81.5, 90.5) | 86.5 | (81.5, 90.5) | 8.3 | (1.8, 22.5) |
Skull fracture | 21/319 | 11.6 | (3.8, 24.6) | 94.2 | (90.7, 96.64) | 86.9 | (82.6, 90.5) | 23.8 | (8.2, 47.2) |
Scalp hematoma | 81/319 | 26 | (14.2, 41.1) | 74.7 | (69.1, 79.8) | 85.7 | (80.6, 89.9) | 14.8 | (7.9, 24.5) |
Neuro deficit | 102/313 | 68.9 | (53.4, 81.8) | 73.5 | (67.8, 78.7) | 93.4 | (89.1, 96.3) | 30.4 | (21.7, 40.3) |
ALOC | 134/316 | 67.4 | (52.0, 80.5) | 61.9 | (55.8, 67.7) | 91.8 | (86.8, 95.3) | 23.1 | (16.3, 31.2) |
Abnormal behavior | 100/314 | 53.3 | (37.9, 68.3) | 71.8 | (66.0, 77.1) | 90.2 | (85.4, 93.8) | 24 | (16.0, 33.6) |
No spontaneous eye opening | 61/316 | 34.1 | (20.5, 49.9) | 83.1 | (78.1, 87.3) | 88.6 | (88.1, 92.3) | 24.6 | (14.5, 37.3) |
Not oriented | 189/305 | 65.1 | (49.1, 79.0) | 66.4 | (60.3, 72.1) | 92.1 | (87.3, 95.5) | 24.1 | (16.7, 33.0) |
Not following commands | 76/310 | 43.2 | (28.34, 59.0) | 78.6 | (73.2, 83.4) | 89.3 | (84.6, 93.0) | 25 | (15.8, 36.3) |
Amnestic | 28/224 | 18.5 | (6.3, 38.1) | 88.3 | (83.0, 92.5) | 88.8 | (83.5, 92.8) | 17.9 | (6.1, 36.9) |
Combined criteria | 271/322 | 95.7 | (85.5, 99.5) | 17.8 | (13.5, 22.9) | 96.1 | (86.5, 99.5) | 16.6 | (12.4, 21.6) |
Abnormal neuro exam | 207/321 | 87 | (73.7, 95.1) | 39.3 | (33.5, 45.3) | 94.7 | (88.9, 98.0) | 19.3 | (14.2, 25.4) |
Abnl neuro exam or vomiting | 217/321 | 87 | (73.7, 95.1) | 36 | (30.3, 42.0) | 94.3 | (88.0, 97.9) | 18.5 | (13.6, 24.4) |
There is substantial evidence in the medical literature that brain herniation secondary to lumbar puncture is extremely uncommon; this is particularly true among patients who do not have high-risk clinical findings, such as altered mentation or focal neurological deficits. Because of concerns about this catastrophic possibility, however, several authors have suggested that head CT scanning be performed prior to LP, to help identify patients at increased risk.
Because of the absence of any prior criterion standard regarding “high-risk” CT findings, we performed Part I of our study to define a group of such findings about which experts would agree that they represent an increased risk of LP-induced brain herniation on head CT findings. Using standard accepted methodology of a modified Delphi technique, our expert panel was able to reach consensus, and at least one of their “high-risk” findings was present in about one of every seven patients in our cohort; a majority of these patients at “high risk” as indicated by their CT findings had clinical findings (as
There are two very different possible explanations for this discordance between clinical findings in individual patients and “high-risk” elements on CT. One explanation is based on the
We believe that this disjunction between sophisticated technology and traditional diagnostic evaluation is not unique to head CT scanning and the risk of herniation and that the importance of our findings extends far beyond the narrow question of whether routine imaging (which also carries well-described medical and economic costs [
Although it is tempting to assume that technology is both more reliable and more accurate than clinical examination, there are several reasons to question this belief. First, there are numerous examples where some finding identified by some type of advanced technology is given the same name as a previously well-known clinical event, but it actually implies a far less dangerous clinical entity. As in the case of PE, described above, a hyphema diagnosed by slit-lamp examination is less worrisome (and should be approached differently) than a hyphema visible to the naked eye; similarly, CT-defined pneumothorax is clinically different than pneumothorax seen on a chest X-ray, acute myocardial infarction (MI) defined on the basis of troponin leak is not the same as the clinically apparent MI, and asymptomatic microscopic prostate “cancer” diagnosed by biopsy, after screening, implies a very different prognosis (and should be approached differently) than does cancer searched for and found because of clinical symptoms.
With regard to the specific clinical issue raised in our study, a recent publication reported head CT “evidence of herniation” in many patients who were clinically entirely stable [
Since clinical brain herniation almost inevitably results in death, it is appropriate that CT criteria for safe performance of LP should err on the side of high sensitivity. On the other hand, failure to recognize that “findings” on head CT may not have the meaning traditionally attributed to the same “abnormalities” (“lesions” and “diseases”) identified on the basis of clinical condition could lead to dramatic overdiagnosis [
In a broader sense, our study raises generic concerns about the danger of
The head CT criteria defined by our panel as representing “increased risk” of herniation are based on expert opinion, rather than experimental evidence, which does not exist. Although many specialties were represented in the expert panel, not all regularly perform LP, which may have resulted in an overly conservative list of “contraindications.” Our expert panel also suffered a degree of expert attrition, with one neurosurgeon, one neurologist, and one neuroradiologist failing to complete the entire sequence of surveys. Peer pressure can influence expert panels [
Nevertheless, our study provides strong evidence that no H&P findings, alone or in combination, are adequately sensitive to detect head CT abnormalities believed by a panel of experts to predict enhanced potential for brain herniation during LP. Since clinical brain herniation is extremely rare following LP and these CT findings are far more common, it is likely that these criteria are overly sensitive and that their application to patients needs to be reconsidered. Furthermore, our study suggests that there is an urgent need to question the assumption that “advanced” technology defines the criterion standard when there is a clear disjunction between abnormalities defined clinically and “abnormalities” given the same name, despite an absence of clinical correlates, when identified by such technology.
Dear Colleague:
Thank you for agreeing to be part of an expert panel to help determine CT contraindications to LP. This is round 1 of 3 questionnaires regarding the topic.
For the following 3 items: Please Please
Your Name The CT finding of a lateral shift of the midline structures is a contraindication to performing an LP. Strongly disagree 0 1 2 3 4 5 Strongly agree Your level of agreement with statement #1 Please explain the reason(s) for your answer: The CT finding of a loss of the basilar cisterns is a contraindication to performing an LP. Strongly disagree 0 1 2 3 4 5 Strongly agree Your level of agreement with statement #2 Please explain the reason(s) for your answer: The CT finding of an obliteration of the fourth ventricle is a contraindication to performing an LP. Strongly disagree 0 1 2 3 4 5 Strongly agree Your level of agreement with statement #3 Please explain the reason(s) for your answer: The CT finding of an obliteration of the superior cerebellar cistern and the quadrigeminal plate cistern with sparing of the ambient cisterns is a contraindication to performing an LP. Strongly disagree 0 1 2 3 4 5 Strongly agree Your level of agreement with statement #4 Please explain the reason(s) for your answer: Please suggest any additional CT findings that are contraindications to LP that you feel should be included in this list, and briefly explain why.
No direct funding was received for this study. The authors were personally salaried by their institutions during the period of writing (although no specific salary was set aside or given for the writing of this paper).
The authors declare that they have no conflict of interests.
William R. Mower, Steven Go, and Kelli N. O'Laughlin conceived and designed the study. Steven Go supervised the expert panel process and data collection. William R. Mower and Guy Merchant supervised the collection and management of data related to head CTs and radiology reports. Kelli N. O'Laughlin, Gelareh Z. Gabayan, Erum Iqbal, and Roberto A. Lopez-Freeman assisted with data collection. Kelli N. O'Laughlin and Gelareh Z. Gabayan categorized radiology reports, and Michael Zucker and Jerome R. Hoffman served as arbitrators. Kelli N. O'Laughlin analyzed the data and William R. Mower served as senior advisor providing guidance on data analysis. Kelli N. O'Laughlin drafted the paper except for the sections related to the expert panel, which were drafted by Steven Go. Jerome R. Hoffman revised the paper (including particularly the Discussion), Gelareh Z. Gabayan participated in paper revision, and Kelli N. O'Laughlin, Jerome R. Hoffman, and William R. Mower took responsibility for the final paper revisions. Kelli N. O'Laughlin, Jerome R. Hoffman, and William R. Mower took responsibility for the paper as a whole.