International Journal of Pancreatology, 3 (1988) 273-282 Elsevier

PAN00108

273

Deaths from acute pancreatitis: why do we miss the diagnosis so frequently?*

Colin Wilson and Clement W. Imrie

Division qf Surgery, Royal Infirmary, Glasgow, U.K.

(Received 27 February 1987; Accepted 29 June 1987)

Key words: Acute pancreatitis; Mortality; Diagnosis

Summary

Review of acute pancreatitis deaths between 1974 and 1984 revealed 126 patients, 53 (42%) of whom were first diagnosed at autopsy. Pancreatitis was undiagnosed in life in 13 (33%) of the patients with a gallstone aetiology, 8 (40%) of those with an alcohol aetiology, 21 (42%) of those of unknown aetiology and 17 (74%) of those with pancreatitis due to other identified aetiological factors. These patients more commonly died within the first week of their illness, more often had significant under- lying medical disease and tended to have less severe degrees of pancreatic parenchy- mal destruction. Only 7 (13%) of the 53 patients undiagnosed in life had presented with abdominal pain. The majority (68%) presented atypically with known or sus- pected medical conditions, often to internists, or post-operatively (19%). Failure to diagnose these patients in life was most often due to a lack of consideration of the diagnosis, usually because of the atypical presentation, and consequent omission of an amylase determination--made in only 5 (9%) of the patients.

Introduction

The reported incidence of deaths from acute pancreatitis which are undiagnosed clinically before autopsy varies widely, ranging from 22% [1] to 86% [2]. Combining

Correspondence to. C.W. Imrie, Consultant Surgeon, Royal Infirmary, Glasgow, G4 0SF, U.K. * Presented in part at the European Pancreatic Club 18th Meeting, Nijmegen, The Netherlands, Septem-

ber 1986, and at the International Association of Pancreatology, Silo Paolo, Brazil, September 1986.

0169-4197/88/$03.50 �9 1988 Elsevier Science Publishers B.V. (Biomedical Division)

274

the quoted figures from seven series [1-7] provides a mean incidence of patients first diagnosed at autopsy of 41.6% of the total number dying from this disease. Previous studies have not analysed this sub-group of patients in any detail, particularly with respect to mode of presentation, reasons for misdiagnosis, clinical course and cause of death. We have therefore undertaken, within a single centre with a particular interest in this disease, a clinical and pathological study of all recorded deaths in which acute pancreatitis was found. A series of prospective clinical trials coincided with this in-depth survey.

Patients and Methods

Patients presenting with acute pancreatitis have been documented prospectively at Glasgow Royal Infirmary since 1971. Between 1974 and 1984 we recorded 73 deaths among 817 patients treated prospectively, a mortality rate of 9%. Some of these patients have been documented in earlier publications [8-11]. Examination of the Hospital Activity Analysis (ICD Code 577.0) over this time revealed 975 patients with a recorded diagnosis of acute pancreatitis, including an additional 53 patients first diagnosed at autopsy whose case records were analysed retrospectively. The total of 126 deaths, representing an overall mortality rate of 13%, forms the basis of this report. The diagnosis of actue pancreatitis was confirmed definitively at autopsy in 79%, at laparotomy in 8% and was based on biochemcial criteria (serum amylase > 1200IU/1) with a compatible clinical picture in the remainder [12].

Pancreatitis was attributed to gallstones when demonstrated at autopsy or lap- arotomy and in two patients on radiological evidence alone. Pancreatitis associated with alcohol abuse has been previously defined, most admitting to an intake well in excess of 80 g of ethanol/day prior to illness [12]. Both those remaining patients with infrequent aetiologies and those with no identified cause for acute pancreatitis have been studied as separate groups. Pancreatic morphology was classified according to the appearance of the pancreas at autopsy or laparotomy, supplemented by histo- logical data where available.

Deaths were categorised as 'pancreatic' when a fatal outcome appeared to have been a direct consequence of acute pancreatitis and included in this group were patients in whom myocardial infarction or pulmonary embolism complicated severe acute pancreatitis. Where death appeared to be a consequence of another medical condition (e.g., liver failure or hypothermia) or where underlying medical disease (e.g., ischaemic heart disease, carcinoma or chronic obstructive airways disease) was a major contributory factor, such deaths were categorised as 'medical'. Statistical analysis was by Student's t-test and the Chi-square test.

275

TABLE I

FREQUENCY OF AMYLASE DETERMINATION MADE IN LIFE IN PATIENTS DIAGNOSED IN LIFE AND AT AUTOPSY

Diagnosed in life Not diagnosed until autopsy

Serum amylase not determined 2 (3%) 48 (91%)

Serum amylase determined 71 (97%) 5 (9%) Amylase normal 1 3 Moderate elevation 2 2 Amylase > 1200 IU/P 68 b 0

Total 73 53

a Phadebas (normal range 70-300 IU/1). b Remainder diagnosed at laparotomy.

Results

The annual incidence of acute pancreatitis was remarkably steady at approx. 80-90 patients per annum. Seventy-five patients (60%) died during the first half of the study compared with 51 in the second half, representing overall mortality rates of 15% and 11%, respectively. Thirty-eight (72%) of the patients first diagnosed at autopsy were recorded in the early part of the study and the fall in the overall mor- tality is partly accounted for by a reduction in the hospital autopsy rate. In contrast, the mortality rate of 9% for those patients diagnosed in life and documented pro- spectively has remained unchanged between the first and second periods of the study.

Diagnosis

Of the 73 patients studied prospectively, 68 (93%) were initially diagnosed on clinical and biochemical evidence and the remainder at laparotomy (Table I). Only 5 of the 53 patients (9%) first diagnosed at autopsy had an amylase value determined in blood or urine during their admission. Thirty-six (68%) of these undiagnosed patients had presented atypically with known or suspected medical diasease, 32 (60%) to internists rather than surgeons (the normal emergency referral pattern). Postop- erative deterioration due to unsuspected pancreatitis accounted for 10 (19%) of these patients and only 7 (13%) had presented with abdominal pain (Table II).

Aetiology of attacks

The age and sex distributions of the various aetiological groups are shown in Table III. Six patients with gallstones had additional aetiotogical factors identified

276

TABLE II

COMPARISON OF PRESENTING SYMPTOMS IN PATIENTS DIAGNOSED IN LIFE AND AT AUTOPSY

Diagnosed in life Not diagnosed until autopsy

Post-operative deterioration 1 (1%) 10 (19%)

Abdominal pain 54 (74%) 7 (13%)

Atypical presentation 18 (25%) 36 (68%) Collapse/confusion/CVA 4 6 Pulmonary 0 6 Cardiac 2 5 Liver failure 1 5 Metabolic (hyperglycaemia) 1 5 Hypothermia 0 4 Abdominal 8 3 Renal 2 2

Total 73 53

(alcohol abuse in three, recent surgery in two and hypothermia in one). Eighteen (69%) of the patients with gallstones who were diagnosed in life had calculi present in the extrahepatic ducts compared with only 5 (38%) of the patients diagnosed at autopsy but this difference was not significant (Chi-square test).

Seven (35%) of the patients with acute pancreatitis related to alcohol abuse had associated alcoholic liver disease. Four of these patients died in frank hepatic failure,

TABLE III

COMPARISON OF AGE, SEX AND AET1OLOGY IN PATIENTS DIAGNOSED IN LIFE AND AT

AUTOPSY

Aetiology Total Diagnosed in life Not diagnosed until autopsy

(%) M:F Mean age (%) M:F Mean age

Gallstones 39* 67 12:14 67 33 7:6 69 Alcohol 20 60 11:1 42 40 5:3 58 a Other causes 23 26 3:3 50 74 9:8 66 b Unknown causes 50 58 12:17 62 42 9:12 62

Total 132" 58 38:35 62 42 30:29 64

* Six patients with > 1 aetiological factor. a p < 0.01, Student's t-test, b p < 0.05, Student's t-test.

277

pancreatitis being first diagnosed in each at autopsy. The mean age of the patients first diagnosed at autopsy was significantly higher than those diagnosed in life (Table III).

Pancreatitis due to identified aetiologies other than gallstones or alcohol was most commonly misdiagnosed, 17 patients (74%) being first diagnosed at autopsy. Ae- tiologies in these patients included post-operative pancreatitis in 11 (following upper abdominal surgery in 8 and cardio-pulmonary bypass surgery in 3), pancreatic car- cinoma in 4, hypothermia in 4 and following translumbar aortography in 2. Ten of the 11 patients with post-operative pancreatitis were first diagnosed at autopsy. The mean age of the patients first diagnosed at autopsy was again significantly older than those diagnosed in life (Table III).

The largest group of patients were those in whom no definite aetiological factor could be implicated despite laparotomy or autopsy in all but 13. Suspected aetiologies in these patients included ischaemia/hypotension in six, septicaemia in three and steroid drug therapy in two. This group included five patients who presented in hy- perglycaemic coma and were diagnosed as having acute pancreatitis at autopsy. Only two of these five patients were known diabetics.

Significant pre-existing medical disease was present in 60% of patients, being least common in patients with gallstone-associated pancreatitis (41%) and most common in those with other identified aetiologies (74%). The incidence of underlying medical disease was similar in three groups of patients whether the diagnosis was made at autopsy or in life, the exception being those with alcohol-associated pancreatitis, where the incidences were 87.5% and 50% respectively. Overall, 66% of the patients first diagnosed at autopsy had underlying medical disease compared with 55% of those diagnosed in life.

TABLE IV

COMPARISON OF TIME OF DEATH AND PANCREATIC MORPHOLOGY IN PATIENTS DI- AGNOSED IN LIFE AND AT AUTOPSY

Pancreatitis Diagnosed Time to death" Not diagnosed Time to death" Total morphology in life (days) until autopsy (days)

Interstitial 9 (12%) 4(1-138) Haemorrhagic 10 (14%) 5(2-26) Necrosis 17 (23%) 5(1 58) Abscess 13 (18%) 31(8 129) Pseudocyst 7 (10%) 40(33-140) Unknown 17 (23%) 6(2 71)

Total 73

13 (25%) 6(1 33) 22 (17%) 20 (38%) 7(1-48) 30 (24%) 15 (29%) 4(1 37) 32 (25%) 5 (9%) 10(2 17) 18 (14%)

- - 7 (6%) - - 1 7 ( 1 3 % )

53 126

a Values are: median (range).

278

<

0 [..

< [ - <

Z <

M

z

�9 Z

<

r ~

Z

F- <

z

o 8

<

f" EL �9 r ~

~ .-~ < <

o

"o

o

0

.i

L~

0

.!

F

%

o

2

.i

o ~ o ;>

279

Analysis of deaths

Sixty-two patients (49%) died within the first seven days of their illness, compris- ing 57% of the patients first diagnosed at autopsy and 44% of those diagnosed in life. The pancreatic morphology and the time to death of the two groups is compared in Table IV. Patients first diagnosed at autopsy tended to have less severe degrees of pancreatic parenchymal destruction. Abscesses and pseudocysts typically occur later in the course of the illness and patients developing these complications survived longer. No relationship was found between the severity of the pancreatic damage and the time to death.

Death was considered to be 'pancreatic' in 56% of patients overall and 'medical' in 32%. Only 17 (13%) of the 126 patients had no laparotomy or autopsy data. Nine of these 17 patients (53%) died within the first seven days predominantly of cardi- opulmonary problems, five of whom had underlying cardiovascular or respiratory disease. Of the eight patients dying after seven days, five succumbed to renal failure, three of these on a background of pre-existing chronic renal failure. Of the 73 patients diagnosed in life, 60% died of 'pancreatic' and 16% of 'medical' cause compared with 49% and 51%, respectively, for the patients first diagnosed at autopsy (Table V). Early 'pancreatic' deaths, usually from fulminant pancreatitis, most commonly resulted from alcohol abuse in younger patients and from pancreatitis of unknown aetiology in older patients. Gallstones most often gave rise to late 'pancreatic' deaths from abscess and/or necrosis with sepsis and multi-organ failure.

Discussion

At 42%, the incidence of acute pancreatitis deaths first diagnosed at autopsy was identical to the mean incidence from the studies quoted above [1-7]. We were sur- prised at this incidence especially as our data derives from a single hospital with a particular interest in this condition and that during this period a series of clinical trials were being conducted when clinical suspicion would be heightened [8-11]. Lack of consideration of the diagnosis of acute pancreatitis because of an atypical clinical presentation, including unexpected post-operative deterioration, lead to a failure of amylase determination in blood or urine in a disturbing 91% of those patients di- agnosed at autopsy. This failure also accounted for the exclusion of such patients from our prospective studies and strongly suggests the need for increased vigilance in diagnosis.

It is impossible, in a retrospective study, to resolve with certainty whether all the patients first diagnosed at autopsy represent true pancreatitis deaths or whether the pancreatic changes may be secondary or represent simply a pre-terminal event. Some have sought to avoid this controversy by studying only patients with severe forms of acute pancreatitis [1,2,6,13] where pancreatitis was clearly the major cause of death.

280

This approach overlooks the fact that circulatory, respiratory and renal insufficiency may occur in acute pancreatitis in the absence of pancreatic necrosis and, particularly in the presence of pre-existing medical disease which affected two-thirds of these patients, might result in death. Others have recognised that death may occur in as- sociation with only mild pathological changes. Shader and Paxton reported a 42% incidence of oedematous pancreatitis in a series of 100 fatal cases, many of these particular patients dying of shock [14]. Trapnell found mild pathological changes a particular feature of patients dying within the first week of illness [4] although this was not apparent from our data (Table IV).

Although the majority of the patients diagnosed in life died of 'pancreatic' causes, more than half the patients diagnosed at autopsy had death attributed to 'medical' causes. The full spectrum of pancreatic damage was represented among the patients dying from 'medical' causes, although milder degrees predominated. In some of these patients it was difficult to determine the contribution of the acute pancreatitis to the patient's death. Unsuspected and undiagnosed pancreatitis was considered likely to have accounted for the presenting symptoms in 66% of the patients first diagnosed at autopsy because of their presentation with coma, collapse, respiratory insufficien- cy, post-operative deterioration etc. In 11 patients (21%) clinically important pan- creatitis probably developed after admission to hospital, and in only seven patients (13 %) was pancreatitis thought to have been an incidental finding at autopsy. Taking into account the patient's age, medical history and underlying pathology(ies) the majority of these undiagnosed patients were considered unsalvagable. However, in 45%, diagnosis of acute pancreatitis in life may have altered management, perhaps allowing survival in some.

Despite a 23% reduction in our hospital catchment population, the annual inci- dence of acute pancreatitis has remained fairly constant. Either more patients are being correctly diagnosed in life, or the incidence is increasing, as other authors have sugested [7,15]. It is disappointing that the mortality rate of 9% for those patients diagnosed in life has remained unchanged despite the advances in investigation and treatment seen over this time. The preponderance of 'pancreatic' deaths among the patients diagnosed in life, particularly in patients with a gallstone or alcohol aetiology suggests further scope for improvement by more effective surgical treatment of com- plications [16,17] and including, perhaps, early endoscopic sphincterotomy for gall- stones [18].

Routine autopsy of all hospital deaths will uncover a significant and unsuspected incidence of deaths associated with acute pancreatitis and clinicians should be aware of this. Only five (9%) of the patients first diagnosed at autopsy had an amylase determination made during their admission and this is a serious indictment of clinical practice. Screening of all emergency admissions and patients following cardiac and upper abdominal surgery, perhaps by urine amylase test-strip [19] might allow di- agnosis of these cases at an earlier stage and possibly a more favourable outcome in some.

281

References

1 Buggy BP, Nostrant TT. Lethal pancreatitis. Am J Gastroenterol 1983;78:810-814. 2 Whalen J, Rush B, Albano E, Lazaro E. Fatal acute pancreatitis. A clinicopathologic analysis. Am J

Surg 1971;121:16-19. 3 Bourke JB. Incidence and mortality of acute pancreatitis. Br Med J 1977;2:1668-1669. 4 Trapnell J.E. The natural history and prognosis of acute pancreatitis. Ann R Coil Surg Engl

1966;38:265-287. 5 Read G, Braganza JM Howat HT, Pancreatitis--a retrospective study. Gut 1976;17:945-952. 6 Peterson L.M. Brooks J.R. Lethal pancreatitis: a diagnostic dilemma. Am J Surg 1979;137:491~496. 7 Corfield AP, Cooper M J, Williamson RCN Acute pancreatitis: a lethal disease of increasing incidence.

Gut 1985;26:724-729. 8 Imrie CW, Blumgart LH. Acute pancreatitis: a prospective study on some factors in mortality. Bull

Soc Int Chir 1975;6:601 603. 9 Imrie CW, Benjamin IS, Ferguson JC, McKay A J, MacKenzie I, O'Neill J, Blumgart LH. A single-

centre double-blind trial of Trasylol therapy in primary acute pancreatitis. Br J Surg 1978;65:337 341. 10 Dickson AP, O'Neill J, Imrie CW. Hyperlipidaemia, alcohol abuse and acute pancreatitis. Br J Surg

1984;71:685-688. 11 Mayer AD, McMahon M J, Corfield AP, Cooper M J, Williamson RCN, Dickson AP, Shearer MG,

Imrie CW. Controlled clinical trial of peritoneal lavage for the treatment of severe acute pancreatitis. New Eng J Med 1985;312:39%404.

12 lmrie CW, Whyte AS. A prospective study of acute pancreatitis. Br J Surg 1975;62:490~494. 13 Thal AP, Perry JF, Egner W. A clinical and morphologic study of forty-two cases of fatal acute

pancreatitis. Surg Gynecol Obstet 1957;105:191-202. 14 Shader AE, Paxton JR. Fatal pancreatitis. Am J Surg 1966;11h369 373. 15 De Bolla AR, Obeid ML. Mortality in acute pancreatitis, Ann R Coll Surg Engl 1984;66:184-186. 16 Bradley EL, Fulenwider JT. Open treatment of pancreatic abscess. Surg Gynecol Obstet

1984;159:509-513. 17 Beget HG, Krautzberger W, Bittner R, Block S, Bfichler M. Results of surgical treatment ofnecrotizing

pancreatitis. World J Surg 1985;9:972-979. 18 Neoptolomos JP, London N, Slater ND, Carr-Locke DL, Fossard DP, Moossa AR. A prospective

study of ERCP and endoscopic sphincterotomy in the diagnosis and treatment of gallstone actue pancreatitis. Arch Surg 1986:121:697-702.

19 Holdsworth PJ, Mayer AD, Wilson DH, Flowers MW, McMahon MJ. A simple screening test for acute pancreatitis. Br J Surg 1984;71:958 959.