Frequently, the major goal in the management of patients with cancer of the pancreas or other upper abdominal organs is palliation. Neurolytic celiac plexus block (NCPB) is commonly used to treat pain of upper abdominal cancer that fails to respond to narcotic analgesics. Many accounts of NCPB using a variety of indications, followup, and techniques [1-24] have appeared since the initial description in 1914 by Kappis . Recent reviews of the efficacy of NCPB have reached conflicting conclusions. Lebovitz and Lefkowitz  affirm the efficacy of NCPB for pancreatic cancer pain, but Sharfman and Walsh  thought that its effectiveness is not yet proven. Thus at present there is no consensus regarding basic questions such as precise indications for NCPB; efficacy of NCPB for pain from pancreatic and other abdominal cancer; duration of analgesic effect of NCPB; and optimal technique and safety of this procedure.
The present study reports a meta-analysis [28-30] of the literature, conducted to make a quantitative assessment of the efficacy and safety of NCPB for treatment of cancer-related pain. Because of inconsistencies in the nature and completeness of data reported by different authors, only a limited array of basic statistical techniques could be justified.
Literature was searched through BRS Colleague and Docline (National Library of Medicine) for the years 1966 to mid-1993. Search terms were "celiac," "neoplasms," and "pain." Full-length reports (i.e., not abstracts) in the English language were selected provided that they described the use of NCPB in two or more patients with cancer pain. Randomized controlled, nonrandomized controlled, and uncontrolled studies, as well as surveys, were screened. Many studies provided information on NCPB in the treatment of malignant and nonmalignant pain. Only studies that allowed us to extract data specifically on patients with malignant pain were selected. Reference citations listed in these reports were also reviewed and included if they met the above criteria.
Since a meta-analysis is valid only if the studies combined are at least comparable in their trial design, we analyzed the results of randomized controlled trials (RCTs) separately from those of other studies. Data from the largest possible number of patients was extracted regarding the type of cancer, pain characteristics, block techniques, analgesic efficacy, and side effects. Combinable efficacy and safety data from non-RCTs, which unfortunately are the main source of NCPB-related available information, were summed separately from that of the RCTs. The summed results were compared to those of the RCTs in order to estimate their validity.
The most reliable estimate of the clinical efficacy of an analgesic intervention would be a RCT comparing the experimental procedure with the traditional, in this case NCPB with oral analgesics. Trials of analgesic drugs generally are randomized because regulatory agencies (e.g., the U.S. Food and Drug Administration) require such studies before approval of the drug for marketing. However, injection of a drug into a nerve plexus can be done without regulatory approval. Whether this is the reason for omitting randomization, only 2 of the otherwise acceptable 24 papers used randomization. One of these  compared the procedure with an analgesic drug, but had only 20 patients, and the other  compared different approaches to the celiac plexus and had no control group. Hence, the only way to obtain an estimate of the success rate of the procedure is to average responses in the reported uncontrolled series. This has been done with full recognition of the fact that the data are not as reliable as they would be if the patients had been selected at random. Publication bias is another potential cause of inflated efficacy, because bad outcomes are less likely to be selected for retrospective reporting than good ones. However, because journals do publish uncontrolled series, they should also present efforts to document efficacy by averaging these different reports.
In an attempt to control for the heterogeneity created by widely different criteria for selection of patients, different techniques of injection, differing criteria for (unblinded) estimates of pain relief, and different criteria for submitting and accepting papers for publication, we have used, where appropriate, a random effects model to determine weighted averages of results because the confidence intervals around the means are enlarged by the between-study heterogeneity.
Various terms relating to outcome such as "good relief," "satisfactory results," and "successful NCPB" were used to describe results of NCPB. To combine these results, three major categories of NCPB efficacy were identified: complete pain relief, partial pain relief, and minimal/no relief. Only studies with results that could be classified in these categories and that specified the time frame at which efficacy was assessed were included in the efficacy analysis. Results were categorized as acute (up to 2 wk) or long-term (2 wk or more). When possible, the long-term effect was further classified to within versus beyond 3 mo from the time of NCPB, and also up to the time of death. Additional analyses included a comparison of short- and long-term NCPB efficacy in patients with pancreatic versus nonpancreatic intraabdominal malignancies, incidence of adverse effects, and subanalyses of NCPB efficacy, adverse effects, and number of repeated blocks performed according to the radiologic technique used (computed tomography, radiography, fluoroscopy, ultrasound, or no radiologic guidance).
Twenty-four papers met inclusion criteria. Two were RCTs [13,24], one was prospective , and 21 were uncontrolled retrospective studies. These studies provided data on 1145 patients Table 1. Twenty-two papers (enrolling 1117 patients) specified the type of cancer for which NCPB was performed Table 1. Sixty-three per cent (707 patients) had pancreatic cancer and the other 37% had nonpancreatic intraabdominal malignancies Table 2. The two RCTs included 71 patients who had pancreatic cancer and underwent NCPB. However, results of these two studies could not be combined since one  compared three different NCPB techniques, and the other  compared NCPB to analgesic treatment.
Although NCPB was performed for analgesia in all patients, pain characteristics are generally poorly specified, and rarely was a multidimensional description of pain provided Table 3. The average duration of pain prior to NCPB ranged from 2.4 to 7 mo. Pain was primarily abdominal, although back and specifically epigastric or hypochondrial pain were also reported. The quality of pain (reported only in two papers) was deep, constant or crampy, boring, gnawing, and accentuated by palpation. Pain intensity prior to block was reported in four papers and was severe in all. "Intractable pain"--persistent despite medical management--as an indication for block was reported in six papers that included 199 patients (18%).
Short-term analgesic efficacy (1-14 days subsequent to the block) in patients with all types of intraabdominal cancers receiving NCPB was extractable from 18 uncontrolled papers Table 4. Success ("excellent," "good," "satisfactory," "marked to complete," or "partial" relief) was reported in 89% of 989 patients. Further analysis of 274 patients with successful results indicated that 59% experienced complete and 41% partial early (2 wk or sooner) pain relief. Longer term outcome (up to 3 mo or beyond) was reported in 273 patients: approximately 90% maintained partial or complete pain relief Table 4.
Both short- and long-term outcomes for patients with pancreatic cancer alone were quite similar to those in patients who underwent NCPB for any type of cancer. For example, 87% of 304 patients in seven studies restricted to patients with pancreatic cancer only [5-7,9-11,21] had a successful short-term outcome. According to the single RCT which provided information on the short-term efficacy of NCPB in 10 patients with pancreatic cancer , all 10 patients reported at least partial pain relief 2 wk later. Long-term (3-10 wk) partial relief was reported in seven and minimal relief in three. In the other RCT , short-term partial to complete pancreatic pain relief was reported in 70%-80% of the patients, and endured beyond 3 mo in 60%-75%.
Data regarding post-NCPB pain relief at the time of death was presented for 53 patients with all types of cancer in six studies [2,3,11,18,19,23]. This merged data indicated that 73% and 92% had partial to complete relief when death occurred within or beyond 3 mo, respectively. Pain relief until death was achieved according to one RCT  in 60%-75% of patients.
The effect of radiologic guidance on outcome was evaluated in 774 patients in 15 non-RCTs. Sixty-eight percent underwent radiologically guided NCPB: 28% by computed tomography, 34% by radiograph, 5% by fluoroscopy, and fewer than 1% by ultrasound, while in 32% no radiologic guidance was used. Short-term outcomes show a high rate (86%-96%) of successful NCPB regardless of the radiologic technique used Table 5. Ultrasound-guided NCPB results were not taken into consideration due to the small number (only seven) of patients. The two RCTs did not allow a formal comparison of outcomes of different radiologic techniques, but separately showed 90% success with radiographic guidance  and 70%-80% with fluoroscopy . The need for repeat NCPB due to lack of effect was reported in 45/616 patients (7%). The incidence of repeat NCPB according to radiologic technique used is summarized in Table 6.
The three most common adverse effects Table 7 were transient: local pain, reported in 96% of patients in two papers, diarrhea in 44% (five papers); and hypotension in 38% (10 studies). It was difficult to correlate the frequency of adverse effects and radiologic technique used due to inconsistency in reporting of adverse effects by different authors. One paper, for example, reported transient local pain in all 124 patients who underwent NCPB . In another , transient local pain was reported in four of nine patients, and in most of the other papers this adverse effect was not reported at all. More severe adverse effects involving 13 of 628 patients (2%) were reported in 13 papers. These included neurologic complications (1%), such as lower extremity weakness and paresthesia, epidural anesthesia, and lumbar puncture. Significant nonneurologic adverse effects occurred in an additional 1% and included pneumothorax, shoulder, chest and pleuritic pain, hiccoughing, and hematuria. No mortality was reported as a result of NCPB. In one RCT  transient local pain was reported in 18% of the patients, and transient diarrhea and hypotension each occurred in approximately one-third of the patients. In that report, more severe adverse effects occurred in 9/61 patients (15%). Of the 10 patients reported in the other RCT , one developed prolonged diarrhea and two transient hypotension.
Although NCPB for the treatment of cancer pain has been the focus of original reports and review articles for many years, the present study differs from previously published literature in that it describes a quantitative synthesis of results from multiple studies. The statistical approach used to analyze data from multiple trials is known as meta-analysis. Meta-analysis is valid only if the studies combined are comparable in trial design, patient demographics, and therapy tested. Since most of the NCPB literature consists of uncontrolled retrospective trials that lack uniformity in precise criteria for patient enrollment, type of outcome reported, and duration of followup, a formal meta-analysis which compares one treatment to another could not be performed. Instead, the uncontrolled studies were separated from the RCTs (in order to avoid the use of studies with different trial design), and a noncomparative quantitative analysis was performed.
Precise, explicit indications for NCPB or repeat NCPB in patients with cancer pain usually are not described. It is a well known procedure for pancreatic cancer, but is also used for palliation of many other types of intraabdominal cancer (37% of reported patients). Most published reports focus on the block technique rather than on the assessment of patients' pain. Crucial information such as pain intensity, duration, location, and quality exists only in a minority of the studies (25% or fewer). Thus, despite our interest in evaluating clinicians' impressions that NCPB is less likely to be successful when tumor extends beyond the innervation of the celiac plexus (e.g., to involve phrenic or somatic afferents), we were unable to correlate outcome with pain characteristics.
Intractable pain as an indication of NCPB was reported only in 17% of the patients. It seems therefore, that although not formally stated, most authors believe that this procedure should not be reserved for a "last resort" when "nothing else works," but should be applied earlier in the course of the illness. Interestingly, in the single RCT that compares analgesic efficacy of NCPB to that of analgesics , the block was performed only in the terminal stage of the patients' lives (mean of 191 days from diagnosis, and 47 days before death). This timing might reflect the fact that the trial was conducted in a terminal oncology palliative care unit.
The short-term success rate of NCPB both according to a large number of retrospective studies and one RCT  is very high (approximately 90%), regardless of the underlying type of cancer. The term "success" indicates partial to complete relief. Since the concomitant use or doses of other analgesics was not always specified, it is difficult to say whether NCPB can achieve satisfactory analgesia when given as a sole treatment, e.g., to substitute for opioids. However, the data suggest that NCPB can at least provide analgesia in addition to that achieved by opioids, and can reduce their consumption [4,9,21,23,24].
The longer term success rate of NCPB was more difficult to assess because extractable data was available only on a much smaller number of patients. It seems, however, that most patients continue to experience partial to complete relief from NCPB until the time of death. This statement, based on data from non-RCTs, is supported by long-term results in the two RCTs, each reporting pain relief in approximately 70% of patients with pancreatic cancer.
Analysis of outcome as a function of radiologic technique fails to show higher rates of success or a lower incidence of adverse effects for radiologically versus nonradiologically guided methods, even when sophisticated methods such as computed tomography scan or ultrasound were used. This uniformity of outcome may imply that NCPB might be performed successfully by experienced physicians even when expensive radiologic equipment is not available. It should be kept in mind, however, that the analysis was based mostly on retrospective studies that were not designed to compare treatments, and that only prospective RCTs can conduct such comparisons in a scientific fashion.
NCPB is a relatively safe procedure. The common adverse effects of diarrhea and hypotension are reported mostly as transient phenomena. Local pain is often regarded as part of the procedure and therefore is not uniformly reported. More significant adverse effects are uncommon.
In summary, although there are major deficits in the literature analyzed, the present analysis suggests that NCPB is likely to have long-lasting analgesic efficacy for pancreatic and other types of intraabdominal cancer. Adverse effects are common, but transient and mild, and severe adverse effects are uncommon. Further RCTs are necessary to compare NCPB to other analgesic treatments, such as systemic or spinal delivery of opioids, and to evaluate whether it may be appropriate to use it at the first appearance of pain with intraabdominal malignancy.
The primary literature searches were undertaken by the National Library of Medicine under the direction of Kristine Scannell and Ione Auston. We thank Dr. Richard Kitz for continued support and encouragement. Miss Evelyn Hall provided expert secretarial assistance.
1. Lieberman RP, Waldman SD. Celiac plexus neurolysis with the modified transaortic approach. Radiology 1990;175:274-6.
2. Ischia S, Luzzani A, Ischia A, Faggion S. A new approach to the neurolytic block of the coeliac plexus: the transaortic technique. Pain 1983;16:333-41.
3. Gorbitz C, Leavens ME. Alcohol block of the celiac plexus for control of upper abdominal pain caused by cancer and pancreatitis. Neurosurgery 1972;34:575-9.
4. Thompson GE, Moore DC, Bridenbaugh LD, Artin RY. Abdominal pain and alcohol celiac plexus nerve block. Anesth Analg 1977;56:1-5.
5. Owitz S, Koppolu S. Celiac plexus block: an overview. Mt Sinai J Med 1983;50:486-90.
6. Brown DL, Bully CK, Quiel EL. Neurolytic celiac plexus block for pancreatic cancer pain. Anesth Analg 1987;66:869-73.
7. Leung JWC, Bowenwright M, Aveling W, et al. Coeliac plexus block for pain in pancreatic cancer and chronic pancreatitis. Br J Surg 1983;70:730-2.
8. Jones J. Coeliac plexus block with alcohol for relief of upper abdominal pain due to cancer. Ann R Coll Surg Engl 1977;59:46-9.
9. Moore DC, Bush WH, Burnett LL. Celiac plexus block: a roentgenographic, anatomic study of technique and spread of solution in patients and corpses. Anesth Analg 1981;60:369-79.
10. Sharp K, Stevens EJ. Improving palliation in pancreatic cancer: intraoperative celiac plexus block for pain relief. South Med J 1991;84:469-71.
11. Matamala AM, Lopez FV, Sanchez JL, Bach LD. Percutaneous anterior approach to the coeliac plexus using ultrasound. Br J Anesth 1989;62:637-40.
12. Moore DC. Celiac (splanchnic) plexus block with alcohol for cancer pain of the upper intra-abdominal viscera. In: International Symposium on Pain of Advanced Cancer. New York: Raven Press, 1979;357-71.
13. Ischia S, Ischia A, Polati E, Finco G. Three posterior percutaneous celiac plexus block techniques. Anesthesiology 1992;76:534-40.
14. Brown DL. A retrospective analysis of neurolytic celiac plexus block for non-pancreatic intra-abdominal cancer pain. Reg Anesth 1989;14:63-5.
15. Squier R, Morrow JS, Roman R. Pain therapy for pancreatic carcinoma with neurolytic celiac plexus block. Conn Med 1989;53:269-71.
16. Ventafridda GV, Caraceni AT, Sbanotto AM, et al. Pain treatment in cancer of the pancreas. Eur J Surg Oncol 1990;16:1-6.
17. Herpels V, Kurdziel JC, Dondelinger RF. Percutaneous CT guided nerve block of the coeliac plexus and splanchnic nerves. Ann Radiol 1988;31:291-6.
18. Filshie J, Golding S, Robbie DS, Husband JE. Unilateral computerized tomography guided celiac plexus block: a technique for pain relief. Anesthesia 1983;38:498-503.
19. Hupert C, Zeig N, Sieglen-Fortunato L, Ciolino RC. Celiac plexus blocks provide pain relief for malignant and benign visceral pain. N J Med 1987;84:573-6.
20. Black A, Dwyer B. Coeliac plexus block. Anaesth Intensive Care 1973;1:315-8.
21. Hegedus V. Relief of pancreatic pain by radiography-guided block. Am J Radiol 1979;133:1101-3.
22. Buy JN, Moss AA, Singler RC. CT guided celiac plexus and splanchnic nerve neurolysis. J Comput Assist Tomogr 1982;6:315-9.
23. Bridenbaugh LD, Moore DC, Campbell DD. Management of upper abdominal cancer pain. JAMA 1964;190:877-80.
24. Mercadante S. Celiac plexus block versus analgesics in pancreatic cancer pain. Pain 1993;52:187-92.
25. Kappis M. Erfahrungen mit localanasthesie bie bauchoperationen. Verh Dtsch Gesellsch Chir 1914;43:87-9.
26. Lebovitz AH, Lefkowitz M. Pain management of pancreatic carcinoma. Pain 1989;36:1-11.
27. Sharfman WH, Walsh TD. Has the analgesic efficacy of neurolytic celiac plexus block been demonstrated in pancreatic cancer pain? Pain 1990;41:267-71.
28. L'Abbe KA, Detsky AS, O'Rourke K. Meta-analysis in clinical research. Ann Intern Med 1987;107:224-33.
29. Sacks HS, Berrier J, Reitman D, et al. Meta-analysis of randomized controlled trials. N Engl J Med 1987;316:450-5.
30. Thacker SB. Meta-analysis. A quantitative approach to research integration. JAMA 1988;259:1685-9.
31. Jacox AK, Carr DB, Payne R, et al. Management of Cancer Pain. Clinical Practice Guideline No. 9. Rockville, MD: Agency for Health Care Policy and Research, Public Health Service, U.S. Department of Health & Human Services; 1994. AHCPR Pub. No. 94-0592.
32. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clin Trials 1986;7:177-88.