Year : 2013  |  Volume : 16  |  Issue : 3  |  Page : 177--179

Challenges in measuring and comparing the risk of an iatrogenic epidural hematoma

Howard Barkan 
 Saybrook University, San Francisco, California, USA

Correspondence Address:
Howard Barkan
Saybrook University, 747 Front Street #3, San Francisco, CA 94111

How to cite this article:
Barkan H. Challenges in measuring and comparing the risk of an iatrogenic epidural hematoma.Ann Card Anaesth 2013;16:177-179

How to cite this URL:
Barkan H. Challenges in measuring and comparing the risk of an iatrogenic epidural hematoma. Ann Card Anaesth [serial online] 2013 [cited 2020 Sep 26 ];16:177-179
Available from:

Full Text

Meta-analyses integrating the results of multiple independent studies allow consistent interpretation of those studies' results, a more precise estimation of any treatment effect and may explicate variation in the results of individual studies. [1] Hemmerling et al., [2] are performing a major function by using meta-analysis to update prior assessments of the risk epidural catheterization and anesthesia carry in cardiac surgery. However, the validity of meta-analyses depends upon the "combinability" of the individual studies and the completeness and lack of bias in the underlying systematic review from which studies are selected for inclusion in the meta-analysis. [3],[4] Further, interpretation of the outcome rates estimated by meta-analysis is strongly influenced by the control group against which those rates are compared, and by the nature of the control group rates with which the meta-analysis generated rates are compared.

Hemmerling et al., ascertained their cases from a multi-pronged review of the published literature. The publications reviewed ranged from conference abstracts through retrospective and prospective non-experimental studies to randomized controlled trials and prior meta-analyses. The potential biases resulting from using published results to estimate population-wide rates are well-established. [4],[5] Hemmerling et al., note that they did not consider differences among the studies they used as the sources. They combined studies despite the frequent warning against combining studies, which used disparate methodology in discussions of meta-analysis methodology. [1],[4],[6],[7] The authors note that increasing familiarity with epidural catheterization is likely to lead to publication of fewer case studies and case series. This in turn would increase the risk of an under-count of the number of cardiac operations in which epidural catheterization was used resulting in an undercount of the denominator for the calculation of the hematoma incidence rate. Their sample of sources would have been more consistent if it had been restricted to sources of a trustworthy count of operations and hematomas from a defined population, e.g., to randomized clinical trials and prospective registries. Selection of studies guided by one of the accepted indices of study quality would have lessened the threat of selection bias: e.g., the Jadad score [8] for randomized clinical trials and the Newcastle-Ottawa score [9] for non-experimental investigations. A sensitivity analysis comparing rates calculated using Hemmerling et al.'s full sample with those calculated using data from such a restricted sample would screen for the existence and estimate the magnitude of any potential sampling bias.

Counting cases of hematoma, the incidence rate's numerator, is also problematic. Both Ho et al., [10] and Ruppen et al., [11] made their estimates using sources that reported no cases of clinically significant epidural hematoma associated with epidural anesthesia in cardiac surgery. The "rule of 3" [12] was used in both studies to estimate the maximum incidence rate consistent with studies having found no cases in their samples. Further, incomplete and/or delayed diagnosis of perioperative epidural hematoma may both reduce the effectiveness of any therapeutic interventions and contribute to an undercount of the incident cases. [13],[14] The most reliable counts of cases of epidural hematomas are produced by uniformly applied computerized record systems and registries using consistent reporting formats. Volk et al., [15] used data reported for 2008 and 2009 to a regional anesthesia registry by all participating clinics to ascertain both the number of neuraxial procedures performed and the number of patients experiencing epidural hematomas. The ascertainment problems discussed above may have biased Hemmerling's ascertainment, explaining why Volk's rate of 1:6,628 for all operations is only marginally lower than Hemmerling's estimated rate of 1/5,493 for high-risk cardiac procedures. Whatever the belief about whether Volk's rate or Hemmerling et al.'s rates are higher, it should be clear that these rates have differing foundations. Volk's rate is based on a direct count of data reported consistently for all those at risk (i.e., all those undergoing operations within defined geographical boundaries and within a delimited time period). In contrast, Hemmerling et al.'s rate is an estimation based on data from various sources with acknowledged unreliability in counts both of those at risk (as Hemmerling noted) and of those developing hematomas. [13],[14]

Hemmerling et al., compare the iatrogenic risk of epidural hematomas with several other risks. Some of these comparisons are straightforward. The risk of a perioperative epidural hematoma is compared with the risks of post-operative acute renal failure following coronary artery bypass grafting, [16] of sternal wound infection [17] and prescribing errors among hospital in-patients. [18] Data for all these comparisons were drawn from records of the hospitalizations during which the events occurred. Numerators and denominators used to calculate these risks are thus parallel to the counts, which are appropriate for calculating the risk of perioperative epidural hematomas. Comparison among these risks is thus appropriate.

Hemmerling et al., also argue that the risk of perioperative epidural hematomas is equivalent to the yearly risks of women developing breast cancer, of men developing prostate cancer, [19] and of individuals of either gender dying in a motor vehicle accident (MVA). [20] All these comparisons are problematic because the risks are calculated for events happening over very different time periods during which subjects were at risk. The onset of the epidural hematoma was within approximately 24 h of epidural catheter placement in all three of the cases presented by Hemmerling et al., In contrast, the incidence rates cited for breast and prostate cancer and the MVA mortality rate are all annual rates. In other words, these incidence rates were calculated for a time period at risk that is approximately 350 times longer than the period during which epidural hematomas occur. In short, for these risks to be describing risks over equivalent periods, the cancer incidence rate and automobile mortality rates should be divided by 350. This transformation to equivalent time periods (i.e., approximately daily) makes the risk of epidural hematoma in the perioperative period approximately 350 times the incidence of breast and prostate cancer and the mortality rate from MVAs Incidentally, the same concern applies to Hemmerling's comparison of epidural hematoma rates with annual heart disease mortality. [21] Again the rate of an event for which the risk period is approximately 1 day is being compared with an annual rate. Straightforward comparison of the risk periods for which the rates were calculated shows that, rather than being one tenth the heart disease mortality rate, adjusting the mortality rate to equivalent time durations finds the epidural hematoma rate to be approximately 35 times the heart disease daily mortality rate.

Estimation of risks and comparisons among risks are best built on solid foundations. Questions can be raised about the reliability of both the numerator and the denominator used to calculate the risk of perioperative epidural hematomas. Questions can also be raised about the comparison of that perioperative risks with annual risks. Risks should be calculated for equivalent time periods before they are compared. Direct comparison of the rate of an event, which can only happen on a single day with an annual rate, is problematic and potentially misleading. Transformation so the risks are for events happening over equivalent time periods erases the apparent and claimed equivalence of some of these rates and reverses the relative magnitudes of others. Meta-analyses have the potential of making important contributions to evaluations of the effectiveness and risks of clinical procedures, but to do so they must avoid tempting methodological short-cuts. Comparison of rates calculated in different ways without explicit caveats about or correction for those differences generates an apparent precision that may be spurious, misleading, or both. [22] Restriction of secondary analysis to data from high quality sources [8],[9] and comparison only among rates evaluated for equivalent periods at risk would both have strengthened this evaluation of an important risk. These steps would also have increased the confidence of clinicians using the perioperative epidural hematoma incidence rate in a risk-benefit analysis of thoracic epidural anesthesia in their own practices.


1Egger M, Smith GD. Meta-analysis. Potentials and promise. BMJ 1997;315:1371-4.
2Hemmerling TM, Cyr S, Terrasini N. Epidural catheterization in cardiac surgery: The 2012 risk assessment. Ann Card Anesth 2013;16:169-77.
3Crowther M, Lim W, Crowther MA. Systematic review and meta-analysis methodology. Blood 2010;116:3140-6.
4Egger M, Smith GD. Bias in location and selection of studies. BMJ 1998;316:61-6.
5Egger M, Ebrahim S, Smith GD. Where now for meta-analysis? Int J Epidemiol 2002;31:1-5.
6Davey Smith G, Egger M, Phillips AN. Meta-analysis. Beyond the grand mean? BMJ 1997;315:1610-4.
7Davey Smith G, Egger M. Meta-analysis. Unresolved issues and future developments. BMJ 1998;316:221-5.
8Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ, et al. Assessing the quality of reports of randomized clinical trials: Is blinding necessary? Control Clin Trials 1996;17:1-12.
9The Newcastle-Ottawa score for non-randomized studies. Available from: [Last accessed on 2010 May 10].
10Ho AM, Chung DC, Joynt GM. Neuraxial blockade and hematoma in cardiac surgery: Estimating the risk of a rare adverse event that has not (yet) occurred. Chest 2000;117:551-5.
11Ruppen W, Derry S, McQuay HJ, Moore RA. Incidence of epidural haematoma and neurological injury in cardiovascular patients with epidural analgesia/anaesthesia: Systematic review and meta-analysis. BMC Anesthesiol 2006;6:10.
12Hanley JA, Lippman-Hand A. If nothing goes wrong, is everything all right? Interpreting zero numerators. JAMA 1983;249:1743-5.
13Christie IW, McCabe S. Major complications of epidural analgesia after surgery: Results of a six-year survey. Anaesthesia 2007;62:335-41.
14Ho AM, Li PT, Karmakar MK. Risk of hematoma after epidural anesthesia and analgesia for cardiac surgery. Anesth Analg 2006;103:1327.
15Volk T, Wolf A, Van Aken H, Bürkle H, Wiebalck A, Steinfeldt T. Incidence of spinal haematoma after epidural puncture: Analysis from the German network for safety in regional anaesthesia. Eur J Anaesthesiol 2012;29:170-6.
16Martinelli SM, Patel UD, Phillips-Bute BG, Milano CA, Archer LE, Stafford-Smith M, et al. Trends in cardiac surgery-associated acute renal failure in the United States: A disproportionate increase after heart transplantation. Ren Fail 2009;31:633-40.
17Lu JC, Grayson AD, Jha P, Srinivasan AK, Fabri BM. Risk factors for sternal wound infection and mid-term survival following coronary artery bypass surgery. Eur J Cardiothorac Surg 2003;23:943-9.
18Lewis PJ, Dornan T, Taylor D, Tully MP, Wass V, Ashcroft DM. Prevalence, incidence and nature of prescribing errors in hospital inpatients: A systematic review. Drug Saf 2009;32:379-89.
19US National Cancer Institute, 2009. Available from: http://www.seer.cancer. gov/statfacts/. [Last cited on 2012 Sep 07].
20Unintentional injury deaths, 2009. Available from: [Last cited on 2012 Sep 25].
21Murphy SL, Xu J, Kochanek KD. Deaths: Final data for 2010. Natl Vital Stat Rep 2013;61:1-167.
22Egger M, Schneider M, Davey Smith G. Spurious precision? Meta-analysis of observational studies. BMJ 1998;316:140-4.