Simplified Admission Criterion for Predicting Severe Complications of Gallstone Pancreatitis FREE

GALLSTONES ARE the most common cause of acute pancreatitis worldwide.¹ Although most patients recover uneventfully, a subgroup go on to develop severe complications. Numerous scoring systems, including those of Ranson and colleagues^{2 - 3} and Imrie and colleagues,^{4 - 6} have been devised in an effort to predict which patients will manifest severe pancreatitis. A modified Imrie (Glasgow) or biliary Ranson score of 3 or greater has been associated with severe pancreatitis. These scoring systems are hampered by the fact that they require measurement of multiple factors over a 48-hour period. In a previous study, we identified 5 variables present on admission to the hospital that predicted the development of complications of severe gallstone pancreatitis (GP) necessitating care in the intensive care unit (ICU).⁷ These factors included the following admission values: white blood cell count of 14.5 × 10⁹/L or more, blood urea nitrogen (BUN) level of 4.3 mmol/L or more (≥12 mg/dL), heart rate of 100 beats per minute or more, serum glucose level of 8.3 mmol/L or more (≥150 mg/dL), and an Acute Physiology and Chronic Health Evaluation (APACHE) II score of 5 or greater. This study prospectively validated these admission criteria in a new group of patients with GP and determined if any of these predictors of severe complications of GP are superior to an APACHE II score of 5 or greater, a modified Imrie score of 3 or greater, or a biliary Ranson score of 3 or greater.

Between March 1, 1997, and March 31, 1999, 92 consecutive patients (77 women and 15 men aged 18 years to 76 years [mean age, 39 years] 77 of whom were Hispanic) with acute pancreatitis of biliary origin seen at Harbor–UCLA Medical Center, Torrance, Calif, were prospectively enrolled in this study after obtaining informed patient consent. The study was approved by the Harbor–UCLA Research and Education Institute Human Subjects Committee. The diagnosis of acute GP was based on the presence of the following: (1) upper abdominal pain and tenderness, (2) an elevated serum amylase level, (3) documented gallstones, and (4) the absence of other factors known to cause pancreatitis (ie, alcoholism, hypercalcemia, hypertriglyceridemia, and medications).⁸ Physiological factors and laboratory data were collected on admission and recorded daily.

Most patients having 1 or more of the 4 simple criterion (ie, white blood cell count, ≥14.5 × 10⁹/L; BUN level level, ≥4.3 mmol/L [≥12 mg/dL]; heart rate, ≥100 beats per minute; or serum glucose level, ≥8.3 mmol/L [≥150 mg/dL]) on admission were placed in the ICU. However, the ultimate decision on ICU placement was at the discretion of the admitting surgeon. Initial management of all patients included fluid resuscitation, bowel rest, parenteral analgesia, and cholecystectomy once the acute phase of pancreatitis resolved. Patients were monitored daily for evidence of severe systemic and local complications, including pulmonary, cardiovascular, infectious, renal, hematologic, neurologic, and gastrointestinal tract. For this study, a complication was considered to be severe if the patient required ICU care beyond the initial 24 hours (Table 1).⁷ Necrotizing pancreatitis was defined as lack of contrast enhancement of a portion of or the entire pancreatic gland on an abdominal computed tomographic (CT) scan, or the presence of necrotic pancreatic tissue on abdominal exploration or autopsy. Sterile necrosis was defined as (1) evidence of pancreatic necrosis on abdominal CT scan or surgical exploration and (2) the absence of clinical and laboratory signs of sepsis, failure to retrieve microorganisms on percutaneous aspiration of pancreatic tissue or peripancreatic fluid, or failure to retrieve microorganisms on surgical exploration, debridement, or autopsy. A pancreatic pseudocyst was defined as a homogeneous low-density collection within a well-defined cavity as shown on a CT scan.

Potential univariate correlates of severe complications were identified using χ² or Fisher exact tests wherever appropriate. Continuous variables were compared with the use of the Wilcoxon rank sum test. The sensitivity, specificity, positive and negative predictive values for each of the 4 single criteria, the APACHE II score of 5 or greater (Table 2),⁹ the modified Imrie score of 3 or greater (Table 3),⁶ and the biliary Ranson score of 3 or greater (Table 4)³ were calculated. Factors found to be significant on univariate analysis were entered into a multivariate logistic regression analysis. P<.05 was considered statistically significant. No adjustment was made for multiple comparisons. All statistical comparisons were performed using the SAS statistical software package (SAS Institute, Cary, NC).

There were 92 patients with documented GP, of whom 77 (84%) were women and 15 (16%) were men. The mean age of these patients was 39 years (age range, 18-76 years). Seventy-seven patients (84%) were Hispanic. The mean admission serum amylase level was 1701 U/L (reference range, 181-2526 U/L). Fourteen patients (15%) had severe systemic complications requiring ICU monitoring and therapy (Table 5). The most frequent systemic complications were pulmonary (11%), cardiac (7%), and infectious (5%), followed by hematologic (2%) and neurologic (2%).

Four patients (4%) had a total of 6 local complications. All of these patients had concomitant systemic complications requiring ICU care. One patient had a pancreatic abscess requiring operative debridement, with associated sepsis and adult respiratory distress syndrome. Another patient had sterile pancreatic necrosis and pulmonary failure requiring prolonged mechanical ventilatory support. Both of these patients later developed pancreatic pseudocysts. The remaining 2 patients had necrotizing pancreatitis associated with fatal multiple system organ failure. These 2 patients were the only deaths in the series (overall mortality, 2%).

On univariate analysis a white blood cell count of 14.5 × 10⁹/L or more (P =.03), a serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL)(P<.001), an APACHE II score of 5 or greater (P =.008), a modified Imrie score of 3 or greater (P<.001), and a biliary Ranson score of 3 or greater (P =.03) were statistically associated with the development of severe complications; whereas a BUN level of 4.3 mmol/L or higher (≥12 mg/dL) and a heart rate of 100 beats per minute or more were not (Table 6). An admission serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) had the highest sensitivity (79%), positive predictive value (41%), and negative predictive value (95%) (Table 7). On multivariate analysis of the 4 criteria, only the serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) was statistically significant (P<.001).

There were 7 patients who had diabetes mellitus and 85 patients who did not. The overall complication rates were 43% (3 of 7 patients) for the patients who had diabetes mellitus and 13% (11 of 85 patients) for the patients who did not. The difference was not statistically significant (P =.13). An analysis was performed to determine whether the presence of diabetes mellitus altered the predictive value of the admission serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL). Among the 26 patients who had an admission serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) were 6 with known preexisting diabetes mellitus and 20 who did not have diabetes mellitus. When all 7 patients with diabetes mellitus were excluded from consideration, the complication rate was 45% (9 of 20 patients) for patients with a serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) and only 3% (2 of 65 patients) for patients without an elevated serum glucose level (P<.001). For these 85 patients who did not have diabetes mellitus, the sensitivity, specificity, and positive and negative predictive values for a serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) were 82%, 85%, 45%, and 97%, respectively. As 6 of the 7 patients with diabetes mellitus had a serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL), the criterion was not predictive in this group.

Early assessment of the severity of pancreatitis is useful, both for prognostication and determination of the need for ICU monitoring and aggressive physiological support. Multiple scoring systems have been devised to assist in this assessment. In 1974, Ranson et al² described a scoring system using 11 objective factors collected within 48 hours of admission to predict the severity of pancreatitis. This study was based on 100 patients with acute pancreatitis, of whom 14 had gallstones and 74 had alcoholism as the orgin. Of 79 patients with fewer than 3 signs, 2 died and 9 were seriously ill. However, of 21 patients with 3 or more signs, 13 died and 7 were seriously ill (P<.001). Based on a later study of 450 patients, with the majority (70%) again of alcoholic origin, Ranson³ revised the criteria for patients with biliary pancreatitis. Although these criteria are most frequently cited for assessment of prognosis in the United States, some authorities believe the number of patients with GP in the study was inadequate to provide an accurate prognostic scoring system.¹⁰

In Europe, Imrie et al⁴ published a study of 134 patients with acute pancreatitis, of which 84 (63%) had biliary stone disease and 51 (38%) were alcohol abusers (1 patient had both biliary disease and alcoholism). From this study, 9 factors were determined to be of prognostic significance in determining the severity of the disease. Later, the number of factors was reduced to 8,^{5 - 6} and subsequently, the predictive value of this modification has been confirmed.^{11 - 13} Other scoring systems that have been recommended include the APACHE II and the multiple system organ failure scores.^{7 ,9 - 10 ,13 - 15}

The above scoring systems have several shortcomings. First, 48 hours are required to complete the data collection for the Imrie and Ranson systems. Not surprisingly, the database for all the factors is often incomplete, Blamey et al,⁶ in 405 episodes of acute pancreatitis, reported complete information in only 36% of the patients. Second, some investigators have suggested that by 48 hours after admission, clinical evaluation alone is equivalent to the scoring systems, making them superfluous.¹⁴ In addition, the APACHE II scoring system is more complicated than either the Ranson or Imrie scoring systems and, therefore, disadvantageous for routine clinical application.¹⁰ Furthermore, the definitions of severe pancreatitis used in the various studies have not been clear or consistent. The Glasgow group defined severe pancreatitis as pancreatitis that "failed to settle."⁵ Ranson et al² defined it as pancreatitis associated with the development of "life-threatening complications or requiring more than 7 days of ICU monitoring." In another study that compared the 3 versions of the Imrie (Glasgow) scoring system, severe pancreatitis was defined as the "development of a major complication requiring more than 20 days hospitalization."¹¹ From these examples, it is apparent that the criteria for defining severe pancreatitis have not been uniform. Since pulmonary complications are most frequent and typically occur early in the course of pancreatitis, it would also be more useful if the prognostication could be made on admission so as to stratify patients at high risk for these and other severe complications.

In this study, we evaluated the ability of 4 previously determined single criterion (ie, a white blood cell count of ≥14.5 × 10⁹/L, a BUN level of ≥4.3 mmol/L [≥12 mg/dL], a heart rate of ≥100 beats per minute, and a serum glucose level of ≥8.3 mmol/L [≥150 mg/dL]) readily available at the time of admission, to predict severe local and systemic complications of GP requiring ICU monitoring and treatment. On multivariate analysis, only the admission serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) was predictive of a subsequent adverse outcome, with a sensitivity of 79%, specificity of 79%, a positive predictive value of 41%, and a negative predictive value of 95%. When patients with diabetes mellitus were excluded, the predictive value of a serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) was further increased. This single, easily measured admission value had a better predictive power than the APACHE II, biliary Ranson, or modified Imrie scoring systems.

Others have previously reported the value of simplified prognostic criteria for acute pancreatitis. In 2 studies, Fan et al^{16 - 17} demonstrated that both an admission serum glucose level of 11.0 mmol/L or higher (>200 mg/dL) as well as a BUN level of 7.4 mmol/L or higher (>20 mg/dL) were comparable to the Imrie multifactoral scoring system for predicting severe complications. Importantly, the criteria for adverse outcome in the studies by Fan et al were very similar to those of the present study. In the earlier study by Fan et al,¹⁷ only 203 (55.7%) of the patients had pancreatitis of biliary origin. The sensitivity and specificity of the combined levels of serum glucose and BUN criteria were only 79% and 67%, respectively.

The reason that an admission serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL) might accurately predict severe and complicated pancreatitis is because, in previously nondiabetic patients, the hyperglycemia may be indicative of severe islet cell damage and dysfunction.¹⁶ Many studies have assessed the effects of acute or chronic pancreatitis on the future development of diabetes mellitus, serum glucose intolerance, and markers of pancreatic endocrine and exocrine dysfunctions (ie, the levels of serum glucose, insulin, glucagon, pancreatic polypeptide, and others). It seems as if both a decreased insulin response and an excess glucagon level have been implicated in the abnormal serum glucose metabolism associated with acute pancreatitis. Seligson et al¹⁸ showed that gross changes of the pancreatic ducts on endoscopic retrograde cholangiopancreatography seen after an acute episode of pancreatitis was associated with deficient B-cell function and decreased serum glucose tolerance. Solomon et al¹⁹ found that acute pancreatitis was characterized by fasting hyperglycemia, hyperglucagonemia, and relative hypoinsulinemia. They also found that at 3 weeks after the acute episode, there was a return to normal of plasma glucose, glucagon, and the primary response of insulin levels; however, the secondary response of insulin remained abnormal. Donowitz et al²⁰ found that basal glucagon levels in patients with acute pancreatitis was 9 times higher than in control subjects, and also found that the glucagon response to alanine was 3 to 4 times greater in patients with acute pancreatitis. Other authors have also shown a higher incidence of diabetes mellitus and glucose intolerance in patients with a history of acute pancreatitis suggesting some permanent pancreatic damage remains.^{21 - 22}

This study demonstrated that a single admission laboratory value, a serum glucose level of 8.3 mmol/L or higher (≥150 mg/dL), was the best single, simple predictor of severe complications of GP requiring ICU care. This value was a better predictor than an APACHE II score of 5 or greater, and a modified Imrie or biliary Ranson score of 3 or greater. In addition to being less cumbersome than standard multifactorial scoring systems, use of this single measurement allows rapid assessment of pancreatitis severity in the emergency department and, thus, facilitates immediate triage to the appropriate level of hospital monitoring and care.

Presented at the 71st Annual Session of the Pacific Coast Surgical Association, San Francisco, Calif, February 21, 2000.

Statistical analysis was performed by Roger Lewis, MD, PhD.

Corresponding author: Christian de Virgilio, MD, Department of Surgery, Harbor–UCLA Medical Center, 1000 W Carson St, Box 25, Torrance, CA 90509.

L. William Traverso, MD, Seattle, Wash: The goal of this study was to quickly identify patients that were headed for a complicated course of GP. Using an institutional review board–approved protocol, they prospectively enrolled 92 consecutively admitted patients who had clinical pancreatitis with documented gallstones. The study occurred during a 2-year period (that is about 1 patient admitted per week). Patients were candidates for direct admission to the ICU if they had any 1 of the following criteria: an elevated white blood cell count, BUN level, serum glucose level, or heart rate. The admitting surgeon then decided if the patient was sick enough to go to the ICU (the authors indicate that most patients were admitted to the ICU). For this study a complication was defined as staying in the ICU beyond 24 hours. A variety of specific complications were observed in 15% of the patients. Note that only 2 patients died in this mainly female Hispanic group.

With this limited number of 92 cases, multivariate analysis was able to show that a serum glucose level of 150 mg/dL or more (≥8.3 mmol/L) at the time of admission was a predictor of a complication. Now what does this mean to the clinician? We cannot use an elevated serum glucose level in patients with diabetes mellitus so the authors excluded 7 patients who were diabetic. The clinician cannot use sensitivity or specificity as the endpoint is already known in these calculations. Therefore, the clinician must use the positive predictive value and the negative predictive value. The authors calculate that, in nondiabetic patients, if the serum glucose level is 150 mg/dL or more, then 45% will go on to a complication; if the serum glucose is less than 150 mg/dL, then only 3% will go on to a complication. This is valuable information and it is simply derived on admission as compared with the more cumbersome APACHE II scores or the 48-hour criteria of Ranson or Imrie.

My questions involve the clinical history and the clinical judgment of the admitting surgeon. This judgment was used to decide if the patient was to be directly admitted to the ICU. Did this decision by itself have any positive predictive value for the development of a complication? Was the time from onset of symptoms to admission recorded and, if this period was prolonged, was it a predictor of complication?

The paper spends a lot of time comparing the predictive value of the 4 simple Harbor Hospital criteria which looks at a prolonged ICU stay as an endpoint and the criteria established by Ranson or Imrie which primarily look at mortality as an endpoint. Remember there were only 2 deaths in the Harbor group of 92 patients. The Ranson criteria were correlated with 32 deaths in 450 patients, mainly in alcohol-induced pancreatitis. Only 16% of the 450 patients had gallstones. The Imrie criteria were developed in a group of patients where only 47 patients had gallstone pancreatitis. It was the study of Larvin and McMahon in 1989 that we should also consider. In a series of 290 attacks of pancreatitis, 43% were due to gallstones. The experienced clinician had a 44% predictive value of predicting severe pancreatitis as defined by organ failure or pancreatic collections. The Harbor criteria of a serum glucose level of 150 mg/dL or more on admission has a remarkably similar predictive value of 45% in nondiabetic patients for a complication of pancreatitis that required an ICU stay of longer than 24 hours.

Although simple, the single admission serum glucose level is probably no better than an experienced clinician assessing the patient. What can we do to increase the clinician's assessment to predict mortality, length of hosptial stay, and need for surgical procedures such as necrosectomy? Ranson went beyond his severity index which he published in 1974 by developing an index based on CT scanning with Emil Balthazar. This CT Severity Index (CTSI) of Balthazar and Ranson was first published in 1990 using just 88 patients with mild to severe acute pancreatitis. We have just presented our series of 268 patients with severe pancreatitis where the CTSI was measured on the index CT scan. If the CTSI was greater than 5, then these patients had an 8 times increased chance of death, a 17 times increased chance of staying in the hospital for longer than 20 days, and a 10 times increased need for necrosectomy.

If I have a Harbor criterion of a high serum glucose level in a nondiabetic patient and if the patient has been rehydrated by an experienced clinician, then I believe the patient should go on to imaging in the form of a baseline index CT scan.

S. Eric Wilson, MD, Orange, Calif: I have 2 questions for Dr Meek. The first is philosophical. Do you think that as far as pancreatitis is concerned, the die is cast, ie, the ultimate severity of pancreatitis is already determined when the patient enters the hospital? This would have to be true to derive meaningful predictive factors. Second, biochemical measurements may change over the first 24 hours, and one could imagine great fluxes in the blood glucose level even in 24 hours over a period of stress, infection, infusions of glucose, and others. But the morphologic features do not change. Did you attempt to correlate the initial CTscan findings with your physiological changes?

Theodore X. O'Connell, MD, Los Angeles, Calif: I have 2 questions for the authors. The first is that perhaps the emphasis of the paper should be turned around, since the negative predictive value is good at 95%, while positive predictive value is only 42%. Perhaps it should be that a serum glucose less than 150 mg/dL is a predictor of good outcomes, ie, these patients need no special care and place the emphasis on the negative predictive value. The second question maybe reflects Dr Wilson's question: is the die already cast on admission? Are there any clinical outcome differences based on the use of this criterion?

Jeffrey E. Doty, MD, San Jose, Calif: It seems to me that the complex Ranson criteria are most valuable when we are comparing one study to another, but in actual clinical practice are not particularly valuable. Mostly what we are doing is deciding which patients go into the ICU. Is there any value for this criteria in determining which patients subsequently get a CT scan and which patients you are starting prophylactic antibiotic therapy for advanced pancreatitis? So my questions are: what are the criteria at Harbor to get a CT scan of a patient with pancreatitis? What are the criteria to admit that patient to the ICU? And third, what are the criteria to start antibiotic therapy for a patient, if there are any?

Lawrence W. Way, MD, San Francisco, Calif: I would like to raise in a slightly more direct a question implicit in Dr Traverso's remarks. Was there a deliberate attempt in this study to assess the power of clinical estimates of severity of an attack? This question stems from evidence suggesting that clinical judgment alone enables an experienced clinician to gauge the severity of pancreatitis just about as well as methodical calculations do (Crit Care Med. 1999;27:2272).

Second, was endoscopic sphincterotomy performed early (ie, within the first few days) in the disease in any patients, and if so, could that have affected these data?

Ronald G. Latimer, MD, Santa Barbara, Calif: It is hard to believe that management of these patients did not have something to do with their complications. I would like the authors to address the issue about endoscopic studies, and ultrasound studies, and to tell us specifically how management may have had a direct effect on the complications.

Dr Stabile: These data represent an extension of our earlier work, which originally grew out of the need to address resource allocation for GP. Gallstone pancreatitis is a very common disease in our hospital where we have a large number of young, Hispanic females with gallstone disease. We admit anywhere from 1 to 3 patients per week with GP, as a result of that high prevalence of gallstone disease. We found that our ICU was literally being overtaxed by admission of these patients and sought to define some simple criteria that could be used at the time of initial evaluation to triage these patients as to the appropriate level of care and monitoring. If a patient were to develop a complication, we could then recognize it early and treat it promptly. If the patient was unlikely to develop a complication, then we would not overuse scarce resources.

The Imrie and the Ranson criteria, although very good for population studies and for predicting on a progressive scale the likelihood of death and serious complications, were not intended primarily for use in determining initial level of care and monitoring.

Drs Traverso, Wilson, and Way questioned the comparability of the criterion that we derived, a serum glucose level of 150 mg/dL or more, to the expertise of an experienced clinician in recognizing severe pancreatitis. That has been studied by Larvin and McMahon and others by means of an observational period, rather than an instantaneous clinical decision. Although we would fully agree that the clinical judgment of an experienced surgeon is perhaps ultimately the best determinant of level of care and intervention, this really is not so easy in the emergency department. The disease is by its nature difficult to predict in clinical course, and that is why so many objective criteria have been defined and experience accumulated in an attempt to have hard criteria beyond clinical acumen to help determine how these patients should be assessed in terms of their initial management.

Drs Traverso and O'Connell commented that the incidence of complications among patients with the high serum glucose level was 45%, whereas only 3% of patients with a low value had complications. This underscores the fact that whatever criterion is used, it must be a fail-safe criterion. We do not want to "under-triage" these patients. If anything, we want to "over-triage" them, such that they will be in a setting where they are going to be safely monitored and observed so that when they do develop complications and begin a downward clinical spiral, this will not be missed or treatment will not be delayed. So although our criterion did have a relatively low positive predictive value, it did have a high negative predictive value and, therefore, does not place patients at risk.

The time course of the patient's disease prior to admission was not considered in this investigation because we wanted a criterion that would be universally applicable. We chose to ignore that issue and just accept the fact that some patients were going to be further along in their course at the time of presentation.

Dr Wilson and others asked whether the clinical die is cast at the time of admission. I believe yes, it is. Many surgical and endoscopic studies have addressed that and data suggest that among patients without concomitant cholangitis the course of pancreatitis is pretty much chiseled in stone at the time of admission. We do not obtain CT scans in all patients but reserve this for patients who are initially very ill or have a deteriorating clinical course.

Dr Doty asked about the administration of antibiotic drugs. We routinely use broad-spectrum antibiotics in patients we believe have necrotizing pancreatitis. That decision is not made simply based on the serum glucose value, but also brings into consideration the evolving clinical examination during the initial 24 to 48 hours and the CT scan that is typically obtained in the patients who seem to be headed toward multiple organ system failure, as evidence of evolving or already established necrotizing pancreatitis.

Drs Latimer and Way asked about endoscopic sphincterotomy and removal of common duct stones. In our practice, we only undertake early biliary intervention, whether it be surgical or endoscopic, in patients who have evidence of concomitant cholangitis. I believe that the Folsch study from Germany is the single best randomized prospective trial examining this issue, and it is the only one that categorically eliminated patients with concomitant cholangitis at the time of presentation with gallstone pancreatitis. That study showed no clinical benefit derived from the performance of early sphincterotomy and stone removal in patients with severe gallstone pancreatitis.