The risk of spinal haematoma following neuraxial anaesthesia or lumbar puncture in thrombocytopenic individuals


Dr Joost van Veen, Sheffield Haemophilia and Thrombosis Centre, Royal Hallamshire Hospital, Glossop Road, Sheffield S10 2JF, UK. E-mail:


Neuraxial anaesthesia is increasingly performed in thrombocytopenic patients at the time of delivery of pregnancy. There is a lack of data regarding the optimum platelet count at which spinal procedures can be safely performed. Reports are often confounded by the presence of other risk factors for spinal haematomata, such as anticoagulants, antiplatelet agents and other acquired or congenital coagulopathies/platelet function defects or rapidly falling platelet counts. In the absence of these additional risk factors, a platelet count of 80 × 109/l is a ‘safe’ count for placing an epidural or spinal anaesthetic and 40 × 109/l is a ‘safe’ count for lumbar puncture. It is likely that lower platelet counts may also be safe but there is insufficient published evidence to make recommendations for lower levels at this stage. For patients with platelet counts of 50–80 × 109/l requiring epidural or spinal anaesthesia and patients with a platelet count 20–40 × 109/l requiring a lumbar puncture, an individual decision based on assessment of risks and benefits should be made.

Epidural and spinal anaesthesia are common techniques with advantages over general anaesthesia. Overall, they are felt to be safe techniques but a feared complication is the occurrence of spinal haematoma. Tryba (1993) estimated the risk of spinal haematoma following epidural anaesthesia at 1:150 000 and 1:220 000 after spinal anaesthesia. Vandermeulen et al (1994) reviewed 18 studies combining 200 000 patients who underwent epidural anaesthesia without any cases of spinal haematoma, whereas Stafford-Smith (1996), in a review including 13 case series involving >850 000 epidurals, identified three haematomas (0·0004%). Similarly, Ruppen et al (2006) found a risk of 1:168 000 of spinal haematoma in obstetric patients receiving epidural anaesthesia. The incidence of spinal haematomata appears to be lower in obstetric patients than in older individuals: Moen et al (2004) found an incidence of 1:200 000 in obstetric patients but 1:3600 in elderly women undergoing total knee replacement. Vandermeulen et al (1994) also reviewed 61 case reports of spinal haematoma after epidural or spinal anaesthesia between 1906 and 1994, the majority of which (39 of 61) were described after 1980. Of these 61 cases, 41 (68%) had evidence of abnormal haemostasis. This was due to heparin in 30 cases and a variety of other causes in the remaining 11. Of these, only four had a reported thrombocytopenia of which one also received heparin and another was a chronic alcoholic. Fifteen were reported not to have any haemostatic abnormalities (although two had a spinal ependymoma and one spina bifida occulta with a vascular tumour) and in five no information on haemostasis was available. The review also suggested that the removal of epidural catheters poses an equal risk to insertion. The situation is similar for lumbar punctures (LPs) but there are no reliable estimates on the risk for spinal hematoma.

Therefore, in many of the reports of spinal haematomata following epidural or spinal anaesthesia, risk factors other than thrombocytopenia were present, being similar to reports of spontaneous spinal haematomata (Groen & Ponssen, 1990). After the introduction of low molecular weight heparin (LMWH) in the USA, nearly 60 spinal haematomata related to neuraxial anaesthesia were reported between 1993 and 1998 (Horlocker et al, 2003) whereas this complication was rare in European reports. The higher incidence was probably related to different dosing regimens of peri-operative LMWH in the USA. Introduction of guidelines reduced the frequency (Horlocker et al, 2003). The American Society for Regional Anaesthesia (ASRA), in its Second Consensus Conference on Neuraxial Anaesthesia and Anticoagulation, discusses in detail the relationship between the occurrence of spinal haematomas and haemostatic abnormalities, particularly those related to anticoagulant and antiplatelet agents (Horlocker et al, 2003). In common with other guidelines, however, they did not discuss the risk associated with thrombocytopenia. A degree of thrombocytopenia is a relatively common occurrence but what constitutes a safe platelet count for these techniques in relation to the occurrence of spinal haematomata is debated. Additionally, different causes of thrombocytopenia may have different bleeding risks. In another review, Douglas (2001) recommended a minimum platelet count of 75 × 109/l for epidural anaesthesia but emphasized the importance of the clinical situation and bleeding history. In particular, in patients with idiopathic thrombocytopenic purpura (ITP) with generally good functioning platelets, a level of 50 × 109/l may be sufficient, whereas in patients with HELLP (haemolysis, elevated liver enzymes, low platelet count) syndrome and rapidly falling platelet counts a higher count may be preferable (Douglas, 2001). A more recent publication (Douglas & Ballem, 2008), suggested a minimum level of 40 × 109/l in patients with ITP in whom the risks of general anaesthesia are high. Similarly, Kam et al (2004) also suggested a minimum platelet count of 50 × 109/l for epidural anaesthesia in parturients with ITP and that the entire clinical situation should be taken into account if epidural anaesthesia is considered, including a rapidly falling platelet count. Gill and Kelton (2000) also suggested a minimum count of 50 × 109/l in ITP patients provided there are no suggestions of platelet dysfunction. Bombeli and Spahn (2004) quoted a minimum platelet count of 50 × 109/l but did not further elaborate on this.

We reviewed the current guidelines, case series and case reports on epidural and spinal anaesthesia as well as LPs in thrombocytopenic patients. Relevant papers were identified by Medline searches for thrombocytopenia, spinal h(a)ematoma, subdural h(a)ematoma, epidural, spinal anaesthesia, regional anaesthesia, neuraxial anaesthesia, LP, spinal puncture, contra indication and guidelines. Other papers were identified by cross checking the references in papers identified above. As there is a clear relationship between anticoagulant use and spinal haematomata for which detailed guidelines exist, we limited the searches to thrombocytopenia only in relation to spinal punctures. Finally, the use of laboratory tests, such as thrombelastography, thrombin generation and others to predict bleeding is the subject of further research and is not discussed in this paper.

Case series in regional anaesthesia

We identified nine single centre case series discussing regional anaesthesia in thrombocytopenic patients; these are summarized in Table I (Beilin et al, 1997; Bernstein et al, 2008; Deruddre et al, 2007; Frenk et al, 2005; Rasmus et al, 1989; Rolbin et al, 1988; Sharma et al, 1999; Waldman et al, 1987; Webert et al, 2003). In all studies the procedures were performed at the reported platelet count given below. Some procedures may have been performed after a platelet transfusion, if necessary, to achieve the reported level, and is included in Table I, if reported. All of the studies but one (Sharma et al, 1999) were retrospective and all assessed the safety of regional anaesthesia in obstetric patients except Waldman et al (1987), who discussed multiple caudal epidural blocks in 19 patients with a platelet count <50 × 109/l and malignancies. The main aim of the one prospective study (Sharma et al, 1999), describing 27 parturients who had epidurals with platelet counts <100 × 109/l out of a cohort of 306 patients (38 with platelets <100 × 109/l), was to evaluate thrombelastography (TEG). Patients with abnormal TEG parameters were excluded from epidural anaesthesia (Sharma et al, 1999). Two series specifically assessed the safety of regional anaesthesia in parturients with immune thrombocytopenia (ITP) only (Deruddre et al, 2007; Webert et al, 2003). Altogether these series described a total of 345 patients with a platelet count <100 × 109/l receiving either epidural or spinal anaesthesia. There were no complications. Of these, 246 had a platelet count between 70 and 100 × 109/l, 24 had a platelet count between 50 and 70 × 109/l and 20 had platelet counts <50 × 109/l. In the remainder of the patients the series do not specify how many patients were in each group. The lowest platelet count for an epidural during delivery was in the series by Rasmus et al (1989) who described 14 epidurals during delivery with platelet counts between 15 and 99 × 109/l. The largest series was by Frenk et al (2005) who described 177 patients with a platelet count <100 × 109/l, of which 170 had regional anaesthesia (platelet count 50–100 × 109/l). The upper limit of the 95% confidence interval (CI) for complications in this series was 1·8% (Frenk et al, 2005).

Table I.   Case series of regional anaesthesia in thrombocytopenic patients.
Reference and study designPatient groupPlatelet count and number of patientsComplications (%)Comments on risk factorsConclusion
  1. TEG, thrombelastography; ITP, idiopathic thrombocytopenic purpura; 95% CI, 95% confidence interval.

(Rasmus et al, 1989)
Retrospective review of 2929 parturients with epidural anaesthesia
Adult delivery14 epidurals
Platelet count 15–99 × 109/l
02 patients with severe pre-eclampsia, 1 with amnionitis and 1 with streptococcal sepsisRegional anaesthesia at platelet counts <100 × 109/l may be safe but individual risk benefit assessment should made
(Beilin et al, 1997)
Retrospective review of:
a) epidurals during delivery
b) patients becoming thrombocytopenic after epidural
Adult deliverya) 30 epidurals with platelet count 69–98 × 109/l
b) 22 epidurals with subsequent platelet count 58–99 × 109/l
0Excluded patients with falling platelet counts and bleedingRegional anaesthesia should not necessarily be withheld when the platelet count is <100 × 109/l
(Rolbin et al, 1988)
Retrospective review 2204 healthy random selected parturients. 104 thrombocytopenic, 61 with epidural, 3 with platelet count <100 × 109/l
Adult delivery61 epidurals with a platelet count <150 × 109/l, 2 with a platelet count 50–74 × 109/l and 1 with a count 75–99 × 109/l0Excluded patients with conditions associated with thrombocytopeniaEpidural anaesthesia is safe if the platelet count exceeds 100 × 109/l in otherwise healthy women and the platelet counts is not falling and there are no associated coagulopathies or platelet dysfunction
(Sharma et al, 1999)
Prospective study of the use of TEG during labour:
a) 52 healthy women
b) 254 with preeclampsia, 38 with platelets <100 × 109/l
Adult delivery27 epidurals in patients with preeclampsia and platelet count <100 × 109/l0Patients with abnormal TEG were excluded from epiduralTEG may be used to assess haemostasis in pre-eclamptic women
(Frenk et al, 2005)
Retrospective chart review of 177 patients with platelet count <100 × 109/l
170 received regional anaesthesia
Included patients with gestational thrombocytopenia, preeclampsia and ITP
Adult delivery153 regional anaesthesia with platelet count 70–100 × 109/l
11 regional anaesthesia with platelet count 60–70 × 109/l
 6 regional anaesthesia with platelet count 50–60 × 109/l
Patients with a platelet count >60 × 109/l had predominantly epidural anaesthesia
0Patients with a platelet count >60 × 109/l had predominantly epidural anaesthesia
Upper limit of 95% CI for complications 1·8%
Need to evaluate the risk-benefit ratio on a case-by-case basis before administering regional anaesthesia to parturients
(Webert et al, 2003)
Retrospective review of 119 deliveries in patients with ITP, 42 with epidural
Adult delivery8 epidurals with platelet count >150 × 109/l
8 epidurals with platelet count 101–150 × 109/l
19 epidurals with platelet count 76–100 × 109/l
6 epidurals with platelet count 50–75 × 109/l
1 epidural with platelet count <50 × 109/l
0Not discussedNo specific comments related to regional anaesthesia
(Bernstein et al, 2008)
Retrospective review of 10·369 births, 131 patients with platelet count <150 × 109/l, 65 had regional anaesthesia
Adult delivery56 patients with platelet count 101–150 × 109/l and regional anaesthesia
9 patients with platelet count 79–99 × 109/l and regional anaesthesia
0Patients with platelet count <70 × 109/l had platelet transfusion
Only 20% of patients with a platelet count <100 × 109/l had regional anaesthesia vs. 50% overall
Parturtients with low platelet counts may be candidates for regional anaesthesia
(Deruddre et al, 2007)
Retrospective review of 10·203 births including 59 patients with ITP
 32 with platelets <100 × 109/l (written in French)
Adult delivery25 episodes of neuraxial anaesthesia and platelet count 88–99 × 109/l0Not discussedRegional anaesthesia should not necessarily be withheld in ITP and platelet count <100 × 109/l
(Waldman et al, 1987)Adult, malignancy19 caudal epidural blocks with platelet count <50 × 109/l0Not discussedCaudal epidural blocks in patients with significant thrombocytopenia are safe

Overall, the studies included patients with a variety of conditions leading to thrombocytopenia, were retrospective and, although some excluded specific patient groups from having regional anaesthesia (abnormal TEG (Sharma et al, 1999), bleeding or rapidly falling platelet counts (Beilin et al, 1997), platelet count <50 or 70 × 109/l (Bernstein et al, 2008; Frenk et al, 2005)), exclusion criteria were not always clear and it is therefore difficult to compare the studies or to make recommendations on the safety of specific platelet counts.

Case series in lumbar punctures

For LPs, seven single centre retrospective series were identified and are summarized in Table II (Breuer et al, 1982; Feusner, 2004; Howard et al, 2000; Kitanovski et al, 2008; Ruell et al, 2007; van Veen et al, 2004; Vavricka et al, 2003). In all studies the procedures were performed at the reported platelet count given below. Some procedures may have been performed after a platelet transfusion, if necessary, to achieve the reported level; this is included in Table II, if reported. Five were in a paediatric population with acute leukaemia (Howard et al, 2000; Feusner, 2004; Kitanovski et al, 2008; Ruell et al, 2007; van Veen et al, 2004). Three of these series describe a total of 1918 LPs in patients with a platelet count <100 × 109/l (Howard et al, 2000; Kitanovski et al, 2008; van Veen et al, 2004), the fourth describes 738 LPs in patients with platelet counts over 30 × 109/l and into the normal range without specifying the number of patients at different platelet counts (Ruell et al, 2007) and the final study described LPs in 163 patients with newly diagnosed Acute Lymphoblastic Leukaemia (ALL) with a median platelet count of 87 × 109/l (47–239 × 109/l) without subdividing patients by platelet count (Feusner, 2004). Based on the studies where patients were grouped by platelet count, 39 LPs were performed at a count <10 × 109/l, 204 at counts between 11 and 20 × 109/l, 817 with a count between 21 and 50 × 109/l and 858 with counts between 51 and 100 × 109/l. There were no bleeding complications in these studies. The majority of these patients were contributed by the series of Howard et al (2000) who described 5223 LPs in children with ALL, of which 941 had a platelet count <50 × 109/l. The upper limits of the 95% CI for complications of the latter series were 1·75% for platelets <20 × 109/l and 0·37% for platelets <50 × 109/l (Howard et al, 2000).

Table II.   Case series of thrombocytopenic patients having lumbar punctures.
Reference and study designPatient groupPlatelet count: number of LPsComplications (%)Comments on risk factorsConclusion
  1. ALL, acute lymphoblastic leukaemia; AMl, acute myeloid leukaemia; NHL, non-Hodgkin lymphoma; CSF, cerebrospinal fluid; 95% CI, 95% confidence interval.

(van Veen et al, 2004)
Retrospective study of 72 LPs in 72 ALL patients
Paediatric<10 × 109/l: 9
11–20 × 109/l: 22
21–50 × 109/l: 41
030 (23%) >10 red cells/μl CSFNo support for prophylactic platelet transfusions
May be useful if platelet count <10 × 109/l
(Ruell et al, 2007)
Retrospective study of 738 LPs in 54 ALL patients
PaediatricPlatelet count ranged from 30 to >91 × 109/l
No numbers given for each group
Suggests they used a cut off for platelet transfusion of 29 × 109/l
065 (9%)>10 red cells/μl CSF.
 30 (4%) >500 red cells/μl CSF
Platelet count of >30 × 109/l safe
 No relationship between traumatic tap and platelet count
(Howard et al, 2000)
Retrospective study of 5223 LPs in ALL patients, 941 with platelet count <50 × 109/l
Paediatric<10 × 109/l: 29
11–20 × 109/l: 170
21–50 × 109/l: 742
51–100 × 109/l: 858
0Traumatic (>500 red cells/high powered field) = 10·8% (n = 548)95% CI platelet count <20 × 109/l = 0–1·75%
 95% CI platelet count <50 × 109/l = 0–0·37%
Prophylactic platelet transfusion is not necessary when platelet counts >10 × 109/l
 No conclusions can be drawn for below this level
(Kitanovski et al, 2008)
Retrospective study in 51 patients (61 LPs) with ALL, AML or NHL (written in Slovenian)
Paediatric<10 × 109/l: 5 of 6 had platelet transfusion
11–20 × 109/l: 7 of 19 had platelet transfusion
21–49 × 109/l: 2 of 36 had platelet transfusion
(Feusner, 2004)
Retrospective report of 163 patients with newly diagnosed ALL
PaediatricMedian platelet count 87 × 109/l (47–239 × 109/l)
No numbers given for each group
011 traumatic LPsNo good supporting data for a safe minimum platelet count
(Vavricka et al, 2003)Retrospective review in 66 AML/ALL patients having 195 LPsAdult20–30 × 109/l: 35
31–50 × 109/l: 40
51–100 × 109/l: 43
>101 × 109/l: 77
Patients with platelet counts <20 × 109/l had a platelet transfusion
0>500 red cells per high powered field in: 6: platelet count 20–30 × 109/l
 4: platelet count 31–50 × 109/l
 3: platelet count 51–100 × 109/l
Trend towards more traumatic taps at low platelet counts
Minimum safe platelet count 20 × 109/l
 95% CI 20–30 × 109/l: 0–10%
 95% CI 31–50 × 109/l:0–8·81%
 95% CI 51–100 × 109/l: 0–8·22
 95% CI >101 × 109/l: 0–1·87
(Breuer et al, 1982)
Retrospective study of 20 cancer patients with a platelet count <20 × 109/l
Adult13 of 20 patients had LP without platelet transfusion2 of 13 found to have spinal subarachnoid haematoma at post mortemNot discussedSuggest a minimum platelet count of 20 × 109/l for LP

The series in adult patients described a total of 131 LPs in patients with platelet counts <100 × 109/l (Breuer et al, 1982; Vavricka et al, 2003). Vavricka et al (2003) described 118 procedures at platelet counts <100 × 109/l and 75 with counts 21–50 × 109/l. There were no complications and the upper limit of 95% CI for complications was 10% for counts between 20 and 30 × 109/l, 8·81% for counts between 31 and 50 × 109/l, 8·22% for counts 51–100 × 109/l and 1·87% for counts >100 × 109/l (Vavricka et al, 2003). The study by Breuer et al, (1982) described 20 LPs at counts below 20 × 109/l, 13 without platelet transfusion of which two were complicated by a spinal subarachnoid haematoma found on autopsy. There was no comment on other possible risk factors (Breuer et al, 1982). The latter study, as well as a study by Edelson et al (1974) (see below) describing eight spinal haematoma’s is frequently quoted as a safe lower limit of 20 × 109/l (Breuer et al, 1982; Edelson et al, 1974).

Three studies commented on a relationship between platelet count and traumatic LP as defined by the number of red cells present per high power field or per microlitre of cerebrospinal fluid (CSF). The largest study was by Howard et al (2002), in a paediatric population, who retrospectively examined 5506 LPs in 965 patients with ALL for modifiable and non modifiable risk factors for traumatic tap (Howard et al, 2002). They reported increased 95% CIs for traumatic taps (>500 red cells per high power field) at platelet counts <100 × 109/l. Other risk factors for traumatic/bloody taps were operator experience, black race, age <1 year, LPs within 2 weeks of each other (particularly if the previous tap was traumatic or bloody or was done at a platelet count <50 × 109/l). They also suggested that, although in routine LPs a count of >10 × 109/l is sufficient for LP in children, platelet transfusions may be necessary at counts <100 × 109/l in the presence of bacteraemia or circulating blasts to prevent meningitis or central nervous system (CNS) disease. In respect to the latter, there are two reports suggesting a relationship between traumatic LPs (>10 red blood cells/μl) in children with lymphoblasts in the CSF sample and an increased relapse rate (Burger et al, 2003; Gajjar et al, 2000).

Vavricka et al (2003) also reported a trend towards more traumatic taps at lower platelet counts in adults and recommended platelet transfusions if counts are <20 × 109/l. (Ruell et al (2007) did not find a relationship between traumatic taps and platelet counts.

Case reports in regional anaesthesia

Limited to case reports reporting on either epidural or single episodes of spinal anaesthesia without any form of anticoagulation, we identified 21 such reports (12 epidurals, nine spinal anaesthesia) at platelet counts <100 × 109/l at the time of puncture or catheter removal (Bailey et al, 1999; Moen et al, 2004; Nguyen et al, 2006; Morisaki et al, 1995; Gustafsson et al, 1988; Chang et al, 2003; Ezri et al, 2002; Harnett et al, 2000; Hew-Wing et al, 1989; Kuczkowski & Benumof, 2002; Landau et al, 2003; Moeller-Bertram et al, 2004; Nafiu et al, 2004; Ozgen et al, 2004; Raft et al, 2005; Sibai et al, 1986; Steer, 1993; Tamakawa & Ogawa, 1998; Wulf et al, 1988; Yuen et al, 1999). Sixteen were for analgesia during delivery (Bailey et al, 1999; Nguyen et al, 2006; Moen et al, 2004; Chang et al, 2003; Ezri et al, 2002; Harnett et al, 2000; Hew-Wing et al, 1989; Kuczkowski & Benumof, 2002; Landau et al, 2003; Moeller-Bertram et al, 2004; Nafiu et al, 2004; Raft et al, 2005; Steer, 1993; Yuen et al, 1999; Sibai et al, 1986). Five of these 16 were complicated by a spinal haematoma (three patients with HELLP syndrome, one patient with pre-eclampsia and one patient with coagulopathy following resuscitation for haemorrhagic shock) (Yuen et al, 1999; Moen et al, 2004; Nguyen et al, 2006; Sibai et al, 1986) and one by an intracranial subdural haematoma in a patient with HELLP syndrome (Ezri et al, 2002). All six had additional risk factors for bleeding: in three cases there was a rapidly falling platelet count (from 99 to 21 × 109/l, 71 to 46 × 109/l and from within normal limits to 16 × 109/l respectively) and more than one attempt to insert the catheter (Ezri et al, 2002; Yuen et al, 1999), in two an otherwise unspecified coagulopathy associated with HELLP syndrome was described (Moen et al, 2004) whereas the last case had a platelet count of 93 × 109/l but a bleeding time of 15 min without further information on coagulation parameters available (Sibai et al, 1986). The last case was part of a cohort of 112 thrombocytopenic patients with HELLP syndrome, of whom 16 had epidural anaesthesia with a platelet count of 83 ± 8 × 109/l (Sibai et al, 1986). The other 10 case studies did not report complications with platelet counts ranging from 2 to 90 × 109/l (eight had platelet counts between 63 and 85 × 109/l, one with platelets of 26 × 109/l, one with a count of 2 × 109/l). Underlying diagnoses included gestational thrombocytopenia in one patient (Landau et al, 2003), ITP in five patients (Bailey et al, 1999; Chang et al, 2003; Hew-Wing et al, 1989; Moeller-Bertram et al, 2004; Steer, 1993), eclampsia in two patients (one with chronic disseminated intravascular coagulation; DIC) (Kuczkowski & Benumof, 2002; Nafiu et al, 2004), one with cryptogenic liver cirrhosis (Harnett et al, 2000) and one patient with familial thrombocytopenia (Raft et al, 2005). Two patients in this group had a prolonged prothrombin time (Harnett et al, 2000; Kuczkowski & Benumof, 2002).

The case reports of regional anaesthesia outside obstetric practice all reported the occurrence of spinal haematomata (Gustafsson et al, 1988; Morisaki et al, 1995; Ozgen et al, 2004; Tamakawa & Ogawa, 1998; Wulf et al, 1988). All had potential risk factors in addition to thrombocytopenia. Rapidly falling platelet counts and prolongation of the prothrombin time were reported by three (Gustafsson et al, 1988; Morisaki et al, 1995; Ozgen et al, 2004), two of which also had liver disease (Gustafsson et al, 1988; Morisaki et al, 1995). A further paper described a patient with liver cirrhosis and platelet count <100 × 109/l without further information (Tamakawa & Ogawa, 1998) and the last study described a patient with acute myeloid leukaemia, prior allogeneic bone marrow transplant and aspergillus pneumonia who was found to have an epidural haematoma on autopsy after having had an epidural catheter placed for pain relief and a platelet count between 10 and 48 × 109/l (Wulf et al, 1988). The minimum platelet count (if reported) in this group was 10 × 109/l. No clear conclusions can be drawn but the reports describing complications are in keeping with suggestions that the safety of epidural anaesthesia in thrombocytopenic patients not only depends on the absolute platelet count, but also on the underlying reason for the thrombocytopenia, how rapidly the platelet count is falling and the presence of a coagulopathy (Douglas & Ballem, 2008; Gill & Kelton, 2000; Kam et al, 2004).

Case reports of LP

After excluding reports of LP in the presence of anticoagulants, 11 reports were found that described 18 LPs in thrombocytopenic patients (platelet count 1–63 × 109/l), complicated by spinal haematomas (Scott et al, 1989; Ayerbe et al, 2005; Edelson et al, 1974; Lee et al, 2007; Mapstone et al, 1983; Pai et al, 2002; Wirtz et al, 2000; Wolcott et al, 1970; Blade et al, 1983; Dunn et al, 1979; Masdeu et al, 1979). An additional three patients without thrombocytopenia that had spinal haematomas were also described in these reports (Lee et al, 2007; Masdeu et al, 1979). Eight of the 18 patients were asymptomatic and found on autopsy (Dunn et al, 1979; Edelson et al, 1974; Masdeu et al, 1979). In all but one case, potential risk factors other than thrombocytopenia were present including CNS disease, rapidly falling platelet counts, DIC or multiple attempts/traumatic LP. The patient without other risk factors had ALL and a platelet count of 26 × 109/l (Ayerbe et al, 2005). A further patient with a platelet count of 63 × 109/l had relapsed ALL after a recent bone marrow transplant (Scott et al, 1989). Recent bone marrow transplantation and intrathecal chemotherapy have been associated with an increased risk of intracranial haematomas (Colosimo et al, 2000).


A total of 17 national and international guidelines from blood transfusion and anaesthetic societies were examined for guidance (American National Red Cross 2007; American Society of Anesthesiologists Task Force on Perioperative Blood Transfusion and Adjuvant Therapies 2006; American Society of Anesthesiologists Task Force on Obstetric Anesthesia 2007; British Committee for Standards in Haematology (BCSH 2003a,b); Gibson et al, 2004; Gogarten et al, 2007; Horlocker et al, 2003; Kwaliteitsinstituut voor de Gezondheidszorg CBO 2004; Layton et al, 2006; Llau et al, 2001; National Health and Medical Research Council Australia 2002; Royal College of Obstetricians and Gynaecologists 2008; Samama et al, 2002; Schiffer et al, 2001; Vandermeulen et al, 2005; Working party on obstetric anesthesia of the Belgian society for regional anesthesia 2002). Of these, only four gave a specific recommendation for epidural anaesthesia (American National Red Cross 2007; BCSH 2003a,b; Samama et al, 2002). The American National Red Cross (2007), the French Society of Anesthesiology (Samama et al, 2002) and the BCSH in guidelines for the management of ITP (BCSH 2003a) suggested a minimum platelet count of 80 × 109/l whereas the BCSH in guidelines for platelet transfusions (BCSH 2003b) suggested a minimum platelet count of 50 × 109/l. Additionally, the Australian and New Zealand College of Anaesthetists (2008), in a special interest paper, suggested a safe platelet count of >100 × 109/l and proposed that a count of >80 × 109/l is safe when there are no risk factors and the platelet count is not falling. The American Society of Anesthesiologists Task Force on Obstetric Anesthesia and the Belgian Association for Regional Anesthesia suggested laboratory tests (including platelet count) in patients at risk of haemorrhagic complications but did not specify a safe platelet count (Working party on obstetric anesthesia of the Belgian society for regional anesthesia 2002; American Society of Anesthesiologists Task Force on Obstetric Anesthesia 2007).

Four guidelines gave suggestions for platelet counts in relation to LPs (American National Red Cross 2007; BCSH 2003b; Gibson et al, 2004; Kwaliteitsinstituut voor de Gezondheidszorg CBO 2004). The BCSH suggested a minimum platelet count of 50 × 109/l in the guidelines for platelet transfusions, the American National Red Cross and the Dutch blood transfusion guidelines suggested a minimum count of 40 × 109/l (American National Red Cross 2007; Kwaliteitsinstituut voor de Gezondheidszorg CBO 2004) and the BCSH transfusion guidelines for neonates and older children suggested a platelet count of 20–40 × 109/l (Gibson et al, 2004). The American Society of Clinical Oncology emphasized the paucity of data in this field and therefore refrained from an evidence-based recommendation (Schiffer et al, 2001).

Current practice in neuraxial anaesthesia

Given the uncertainty of safe levels of platelet counts for spinal puncture, it is not surprising that current practice in relation to giving neuraxial anaesthesia in thrombocytopenic patients is highly variable.

Stamer et al (2007) mailed questionnaires to 918 German departments of anaesthesiology. Three hundred and ninety-seven replied, representing 41·9% of deliveries in Germany. More than half of the respondents never performed spinal or epidural anaesthesia when the platelet count fell below 65 × 109/l. For a platelet count of 79 × 109/l, epidural anaesthesia was thought to be contraindicated by 37% and spinal anaesthesia by 22·2% (P = 0·001). There was a greater reluctance to use regional blockade in departments with <500 deliveries/year than in departments with >1000 deliveries/year. Preeclampsia (severity not specified) was considered an absolute contraindication for regional block by 15% and placenta praevia by 30% of respondents. The authors emphasized the need for guidelines (Stamer et al, 2007).

Beilin et al (1996) mailed questionnaires to 153 directors of obstetric anaesthesia in academic centres and 153 private anaesthesiologists. These authors found that 60% of American anaesthetists performed a block at platelet counts between 80 and 100 × 109/l without further investigations in otherwise healthy women during delivery and 16% carried out the procedure at platelet counts between 50 and 79 × 109/l.

Spinal cord injury associated with anaesthesia was a leading cause for claims of nerve damage in the 1990s (Cheney et al, 1999). Cheney et al (1999) identified 73 claims related to spinal cord injury, 50 of which were related to regional anaesthesia (35 lumbar epidurals, nine subarachnoid blocks and four thoracic epidural blocks) with epidural haematoma being the most common reason for injury (16 cases). Major factors associated with spinal cord injury were blocks for chronic pain management (14 claims) and systemic anticoagulation in the presence of neuraxial block (13 claims). They also reported that delays in diagnosis of spinal haematoma were often due to attributing the postoperative weakness or numbness to persisting local anaesthetic effects rather than spinal cord ischaemia. Delays in the recognition of and response to neurological compromise was also emphasized in a recent document of the Victorian Consultative Council on Anaesthetic Mortality and Morbidity (Hughes, 2005).


Although spinal haematomas after spinal puncture are rare, the consequences can be devastating and they are the most common reason for claims resulting from nerve damage in this setting. An extensive review only identified 613 case reports over a 170-year time span (Kreppel et al, 2003). In this report no definite triggering factor could be identified in nearly half of cases (43·6%) but 16·9% of patients were receiving anticoagulant therapy. Spinal/epidural anaesthesia or LP in combination with a haemorrhagic diathesis (mainly anticoagulant therapy) was the fifth common cause (6·0%) whereas spinal/epidural anaesthesia or LP without a haemorrhagic diathesis was a triggering factor in 4·2%. There was no specific mention of thrombocytopenia (Kreppel et al, 2003). Although it is logical that thrombocytopenia will be a risk factor for occurrence of spinal haematomata, the threshold at which this occurs may be fluid and dependent on the co-existence of other risk factors and it is therefore not surprising that no clear guidelines on the safety of these procedures in the presence of thrombocytopenia exist and that clinical practice varies substantially between centres. Previously, platelet counts >100 × 109/l were suggested prior to epidural anaesthesia, primarily based on personal opinion (Bromage, 1993). On current practice, the majority of anaesthetists however would not view a platelet count of 80–100 × 109/l as a contraindication. The same threshold is also reflected in the available guidelines and through the retrospective case series that describe a total of 246 patients having an epidural without complications. Whether regional anaesthesia is safe at counts between 50 and 80 × 109/l is even more difficult to answer. There are studies suggesting that this is indeed safe, particularly in patients with ITP. However, in patients with a rapidly falling platelet count, conditions associated with platelet dysfunction or coagulopathies, as well as in patients where a difficult or traumatic puncture is more likely (such as ankylosing spondylitis), more caution is required. This is also reflected in the case reports where all haematomas in obstetric patients occurred with other risk factors present. An obvious coagulopathy was present in one of the three case reports outside the setting of obstetric practice whereas in the other two the clinical setting makes a coagulopathy a likely possibility (liver cirrhosis and neutropenic sepsis with aspergillus pneumonia) but no coagulation parameters are available.

For LPs there is equal uncertainty regarding a safe threshold of platelet count. In total there were 817 LP’s described in children with a platelet count between 21 and 50 × 109/l and 243 LP’s in children with a platelet count of <20 × 109/l. The vast majority, however, was contributed by the study of Howard et al (2000) and even though their safety record is excellent, with upper limits of 95% CIs of 1·75% and 0·35% for platelet counts <20 × 109/l and 50 × 109/l respectively, it is difficult to give evidence-based recommendations on (predominantly) a single centre retrospective study. Even less evidence is available for adults. The case reports all reported the occurrence of spinal haematomas. This occurred in 11 LPs with a platelet count <20 × 109/l, six LPs with a count between 20 and 50 × 109/l and in four with a platelet count >50 × 109/l (including in two with a normal platelet count). All but one however had other potential risk factors for bleeding. Therefore it may well be safe to perform LPs at platelet counts <20–50 × 109/l (recommended as a minimum platelet count in different guidelines) but, similar to the discussion on neuraxial anaesthesia, this is probably strongly influenced by the presence of other risk factors.


In view of the above we conclude that 80 × 109/l is a safe count for placing/removing an epidural or spinal anaesthetic and 40 × 109/l is a safe count for LP. This, however, is provided that:

  • 1 The platelet count is stable.
  • 2 There is no other acquired or congenital coagulopathy.
  • 3 The platelet function is normal and the patient is not on an antiplatelet drug.
  • 4 The patient is not on an anticoagulant. If the patient is on a low molecular weight heparin, 12 h should have elapsed from the last dose of a prophylactic dose or 24 h after a therapeutic dose before an epidural or spinal anaesthetic is placed (Horlocker et al, 2003).

It is possible that lower platelet counts may also be safe but there is insufficient published evidence to make recommendations for lower levels at this stage. For patients with platelet counts of 50–80 × 109/l requiring epidural or spinal anaesthesia and patients with a platelet count 20–40 × 109/l requiring a LP, an individual decision based on risks and benefits should be made.