EA, using LA, provides analgesia quality superior to that achievable with systemic analgesics such as opioids [4]; it is especially effective at reducing pain during movement, including respiratory movement. By preventing nociceptive impulses from reaching the spinal cord, it may prevent the initiation of intraspinal nerve conduction pathway remodelling found in chronic pain syndromes [17, 18].

In published studies, EA has been associated with interim benefits: (1) reduced pain scores when compared to systemic analgesia [4]; (2) reduction in requirements for opioids for post-operative pain, with reduced opioid adverse effects [18]; (3) decreased rate of graft occlusion after lower-limb peripheral vascular disease surgery [19]; (4) thoracic EA (TEA) reduces ventilated days post abdominal aortic aneurysm surgery [20]; TEA reduced pain and respiratory dysfunction in patients with multiple fractured ribs [21] and (5) TEA showed to be beneficial as therapeutic sympathectomy for cardiac surgery [22].

TEA is the central anaesthetic technique adopted in several multidisciplinary “fast-track” clinical pathways for oesophagectomy, colonic surgery and upper abdominal surgery, which have been associated with shorter duration of ICU and hospital stay [23, 24]. TEA is also useful in post-thoracotomy pain reduction programmes [18].

Safe proficiency in EA placement, especially Thoracic Epidural Analgesia (TEA), requires considerable training. Vertebral anomalies occasionally preclude EA placement, even by skilled practitioners. Inadvertent dural puncture leads to cerebrospinal fluid (CSF) leak and debilitating headache, and spinal cord injury may rarely occur with dural puncture during TEA. Transient neuropathy can also occur, albeit with a low risk of permanent injury. Catheters should be securely fixed to skin to prevent movement, but catheter migration can still occur, usually back out of the ES causing EA failure, although instances of late dural penetration have been documented. It is imperative to ensure that the volumes of LA used in EA are not inadvertently infused into the subarachnoid space and CSF.

EA is contraindicated by local sepsis at the planned insertion site, and is discouraged in the presence of systemic sepsis. Catheter infection from bacterial migration from the insertion site and from unsterile injection technique is possible.

Unpredictable variation in ES anatomy (blood vessels, fat, fibrous septae) limits predictable consistency of both catheter sitting and LA spread. An isolated SNR may remain unblocked due to local anatomical factors despite good LA spread to surrounding segments. These patient factors, in combination with the ever-present possibility of catheter migration, mean that a proportion of EA will fail for technical reasons, and a ‘backup’ analgesic strategy must always be planned. Even when blocks initially work well, repeated clinical assessment of nociceptive sensory dermatome level (using cold stimulus, which correlates with nociception) and overall efficacy of block is required to titrate appropriate continuation dosing. This can be labour-intensive, and requires trained staff.

Epidural blood vessels are at equal risk for injury during catheter insertion and withdrawal. An expanding epidural haematoma will compress nerve roots or spinal cord and may lead to paraplegia. Clinical signs are subtle, and MRI confirmation and surgical decompression within 8 h are essential for there to be a chance of avoidance of permanent neurological fallout. EA is thus contraindicated in patients with significant coagulopathy or thrombocytopaenia. Any catheter manipulation must be temporally separated from the administration of therapeutic anticoagulants; current local consensus guidelines as to anticoagulant timing for patients requiring EA must be followed [25].

Thoracic and lumbar SNRs are accompanied by the sympathetic nerves, which are invariably blocked by even low-concentration LA; hence, a sympathectomy is an unavoidable consequence of EA. The wider the segmental band of analgesia, the more extensive is the sympathectomy, which additionally extends two to three segmental levels higher than the level of sensory block. It is occasionally therapeutically useful: cardiac sympathectomy by thoracic EA may reduce heart rate and arrhythmias after ischaemic heart disease surgery [22], EA reduces graft failure after peripheral vascular surgery [18], microvascular perfusion and function of the gut may be improved [25]. Frequently, however, sympathectomy-induced venodilation reduces cardiac preload, causing hypotension. Bradycardias may also occur. Significant hypotension may result in patients being given more fluids, more blood products and vasopressors, all with attributable risks that may outweigh the benefits of the EA. The hypotension may be especially difficult to manage in patients with other coexisting causes of hypotension such as systemic inflammatory response.

Assessing the role of EA as a general management strategy in the intensive care unit is complicated. Studies specifically evaluating RA of any form in critically ill patients are rare [26, 27]; thus, extrapolations from studies in the overall surgical population must be made. Few studies of EA in the overall surgical population evaluate mortality as a primary endpoint, and many are underpowered. The weight placed by prominent recent meta-analyses on intra-operative events complicates extrapolation of their results to the post-operative period relevant to intensive care [28].

The large randomised controlled MASTER trial, assessing EA (447 patients) versus systemic analgesia (441 patients) in high-risk patients undergoing major abdominal surgery or oesophagectomy with mortality as a primary endpoint, failed to demonstrate mortality benefit (5.1 % EA, 4.3 % control, p = 0.67) even on subgroup analysis [29], although EA patients had lower pain scores and lower respiratory failure incidence (23 % EA, 30 % control, p = 0.02) [30]. The 2013 ACCCM Pain, Agitation and Delirium guidelines were able to recommend thoracic EA for patients after abdominal aortic aneurysm surgery, but found inadequate, or conflicting, data in other areas; hence, no recommendation could be made for thoracic EA in non-vascular abdominal or thoracic procedures, or for lumbar EA [31]. A 2014 meta-analysis, usefully focusing on the post-operative value of EA when compared to systemic analgesia, assessed 9044 patients from 125 studies over 42 years. The authors claimed to demonstrate a small mortality benefit for EA when compared to systemic analgesia only (variably reported as 3.1 % vs 4.9 %, 2 % vs 3.2 %, and 1.8 % vs 2.4 %, depending on the trial inclusion criteria), and showed that EA was associated with reduced Odds Ratios less than 1 for atrial fibrillation, supraventricular tachycardia, respiratory depression, atelectasis, pneumonia, ileus and post-operative nausea, but also confirming substantially increased Odds Ratios for hypotension in EA. This study’s four-decade inclusion period has drawn criticism, but it supports relative safety of EA, a trend to respiratory benefit, confirms the risk of hypotension and suggests a small mortality benefit (of the order of 2 % for EA vs 3 %) [32]. Substantiating these apparently small mortality differences by randomised controlled trials will require an enrolment in excess of 8000 patients [30, 32], which may never be practically feasible.