Improved Prediction of Outcome in Patients With Severe Acute Pancreatitis by the APACHE II Score at 48 Hours After Hospital Admission Compared With the APACHE II Score at Admission FREE

THE CLINICAL course of severe acute pancreatitis is frequently unpredictable and highly variable.^1- 2 Multiple prognostic scoring systems have been developed to discriminate between patients with mild acute pancreatitis and those at high risk for developing severe pancreatitis.^3- 5 Ranson criteria and the admission APACHE (Acute Physiology and Chronic Health Evaluation) II score have been widely used in clinical studies. The APACHE II score is preferable by some researchers^1- 2,6- 8 over Ranson criteria because it can be determined at hospital admission and its measurement is routine for patient examination in many intensive care units (ICUs). These scoring systems may allow early identification of patients with severe pancreatitis, so that appropriate care is provided to this group of ill patients. In addition, the scoring systems allow comparison of outcome of severe pancreatitis between different centers, and are useful for stratifying patients in clinical trials.

Previous studies on prognostic scoring systems have focused on identifying patients with severe pancreatitis from a cohort with mild and severe pancreatitis; most of the patients in these studies had mild pancreatitis. To our knowledge, the role of prognostic systems in identifying patients with a poor outcome in a preselected group with severe pancreatitis has not been examined. This is an important paradigm shift because outcomes from severe pancreatitis have improved during the past decade and, therefore, there is a need to better define the subgroup of patients with severe pancreatitis who do poorly. This study evaluates the APACHE II score at predicting the presence of pancreatic necrosis, organ failure, and patient mortality in a large cohort of patients with severe pancreatitis.

The medical records of patients with a confirmed diagnosis of acute pancreatitis who were admitted to The Los Angeles County–University of Southern California Medical Center, Los Angeles, between December 1, 1996, and August 31, 1999, were reviewed. One hundred twenty-five patients, who demonstrated 3 or more Ranson criteria or remained in the hospital for longer than 6 days for the management of acute pancreatitis and its complications, were included in this study. An arbitrary cutoff of 7 hospital days was used to exclude patients with mild episodes of acute pancreatitis. In addition, patients remaining in the hospital for elective cholecystectomy after resolution of acute pancreatitis were excluded. Of the 125 patients included in the study, 41 (33%) had 2 Ranson criteria and the remaining 84 (67%) had 3 or more Ranson criteria. The Institutional Review Board of The University of Southern California approved the study.

Significant pancreatic necrosis was defined as the presence of nonenhancing pancreatic tissue involving 30% or more of the pancreas as evaluated by a dynamic contrast-enhanced abdominal computed tomographic (CT) scan. The Atlanta definition of severe acute pancreatitis was used.⁹ This classification defines severe acute pancreatitis as the presence of 3 or more Ranson criteria, an admission APACHE II score of greater than 7, the presence of 30% or greater pancreatic necrosis, and the development of an acute pseudocyst, a pancreatic abscess, and/or organ failure. Organ failure, acute pseudocyst, and pancreatic abscess were also defined in accordance with the criteria of the Atlanta symposium.⁹ Demographic data; duration of symptoms at presentation; length of stay; Ranson criteria; cause of acute pancreatitis; presence of pancreatic necrosis, acute pseudocyst, or pancreatic abscess by contrast-enhanced abdominal CT scan; surgical procedures; development of organ failure; patient outcome; and the APACHE II score at hospital admission and at 48 hours after admission were recorded and analyzed by a retrospective medical record review.

Normally distributed continuous variables were expressed as mean (range), and nonnormally distributed variables were expressed as median (range). Continuous nonnormally distributed variables between 2 groups were compared with the 2-tailed, unpaired, Wilcoxon rank sum test. P<.05 was considered statistically significant. Sensitivity, specificity, and positive and negative predictive values, including accuracy, were determined for the APACHE II scores at cutoff points that gave the best combination of sensitivity and specificity.

Of the 125 patients included in the study, 80 (64%) were males and 45 (36%) were females. Their mean age was 40.8 years (range, 14-88 years). One hundred six patients (85%) were Latin American. Alcohol and gallstones were identified as the most common causes of acute pancreatitis. A list of all causes is as follows:

The median duration of symptoms before hospital admission was 1 day (range, 0-7 days). The median number of Ranson criteria was 3 (range, 2-8). The median length of stay was 19 days (range, 2-185 days). Fifty-one patients (41%) were admitted to the ICU.

One hundred six patients (85%) met the criteria for severe acute pancreatitis. Thirty-two (34%) of 93 patients who underwent CT scanning had 30% or greater pancreatic necrosis. Of these 32 patients, 10 (31%) underwent surgical necrosectomy. Organ failure developed in 65 (52%) of the patients; 15 (12%) developed multisystem organ failure (≥3 organs involved), 35 (28%) had single organ failure, and the remaining 15 (12%) had 2 organs involved. Twenty-five (78%) of the 32 patients with pancreatic necrosis developed organ failure. There were 14 deaths (11%). Of these, 9 patients (64% of all deaths) had pancreatic necrosis. The mortality among patients with pancreatic necrosis was 28% (9 of 32 patients); and among patients with organ failure, 22% (14 of 65 patients). The mean hospital stay at the time of death was 25.1 days (range, 2-89 days).

The median APACHE II score at hospital admission and at 48 hours after admission was 6 (range, 0-26). For the whole group, there was no statistically significant (P = .41) change in the APACHE II score during the first 48 hours after admission. The median combined APACHE II score was 12 (range, 1-45). Fifty-one patients (41%) demonstrated an increase in the APACHE II score during the initial 48 hours, while 74 (59%) showed a decrease or no change in the score. There was no significant difference in the admission APACHE II score between the groups that did and did not develop necrosis. On the other hand, we found a significant association between the 48-hour score and the presence of pancreatic necrosis (Table 1). Patients with pancreatic necrosis showed a median increase of 2.5 points in their APACHE II scores over 48 hours, while those without necrosis showed a median decrease of 1 point (Table 1). Similarly, the combined score was significantly higher among the group with necrosis compared with the group without necrosis (Table 1).

The median APACHE II score at hospital admission between survivors and nonsurvivors was not significantly different. By contrast, the APACHE II score at 48 hours after admission was significantly higher among nonsurvivors than survivors (Table 2). Patients with fatal outcomes demonstrated a median increase of 8 points in their APACHE II score after 48 hours; those who survived showed a median decrease of 1 point. This difference was statistically significant (Table 2). The combined score was also significantly higher among nonsurvivors than survivors (Table 2).

Figure 1 shows the individual APACHE II scores at hospital admission and at 48 hours after admission for the 14 patient deaths. In 2 patients, 48-hour data were not available because of deaths within a few hours after admission; in 1, the 48-hour data were not recorded. Of 11 deaths with a 48-hour APACHE II score, 8 showed deterioration in their scores over 48 hours after admission. One patient with a high admission score of 23 had a change to 22 at 48 hours. In only 2 patients, who subsequently died at 28 and 91 days after admission, was a significant improvement of the APACHE II score seen at 48 hours. These results suggest that deteriorating APACHE II scores may identify a subgroup of patients with a poor outcome.

The admission and 48-hour scores displayed a significant association with the development of organ failure (Table 3). Of the 3 major outcome measures studied, a significant association for the admission APACHE II score was found only for organ failure. The combined APACHE II score was also significantly higher in the patients who developed organ failure compared with those who did not (Table 3).

Table 4 shows the significance of the admission, 48-hour, and combined APACHE II scores at predicting death as a measure of patient outcome. The admission APACHE II score accurately predicted patient outcome in only 75% of patients compared with an accuracy of 93% for the 48-hour and combined scores.

The APACHE II score was initially devised as a prognostic scoring system in critically ill patients requiring ICU care.⁴ The APACHE II score provides an immediate physiologic assessment of individual patients, in conjunction with their age and comorbidity. Some researchers^2,6- 7 have preferred the APACHE II score to other prognostic scoring systems for severe pancreatitis, such as Ranson criteria, because the APACHE II score can be determined at hospital admission. For example, in a study of 290 attacks of acute pancreatitis, only 59 (20%) of which were classified as severe, Larvin and McMahon⁶ found the APACHE II score at admission to be accurate at identifying patients with severe acute pancreatitis. In contrast, we found that the admission APACHE II score was not accurate at predicting pancreatic necrosis, organ failure, and death, thus mitigating the importance of its measurement at hospital admission. Our findings seem to be consistent with the understanding of acute pancreatitis as a disease process in evolution, during which worsening physiologic assessment scores, despite appropriate resuscitative measures, are likely to identify patients at high risk for developing complications or a fatal outcome.

Several different prognostic systems have been developed during the past decade, since the initial contribution by Ranson in 1974. There are 3 problems with these studies. First, differing end points and vague descriptions of severe pancreatitis were used. For example, Meek et al¹⁰ define pancreatitis as severe if the patient required ICU care beyond the initial 24 hours, Larvin and McMahon⁶ described severe pancreatitis as the presence of major organ failure and/or a pancreatic collection, and Wilson et al² defined severe pancreatitis as the occurrence of a local pancreatic complication or a systemic complication such as renal or respiratory failure. Second, most studies have small numbers of patients with severe pancreatitis, with most patients having mild pancreatitis. As a consequence, the prognostic accuracy of severe pancreatitis has varied in these studies (Table 5). Third, the practical value of Ranson criteria and APACHE II scores in daily clinical practice is unclear. The complexity of these prognostic scoring systems has precluded routine clinical use in triaging patients because most (85%) have mild pancreatitis. Attempts to define simpler measurements of outcome, such as glucose level¹⁰ or a glucose-urea ratio, have not been successful.¹¹ Furthermore, a comparison of the prognostic criteria with clinical assessment has failed to demonstrate an advantage of the scoring systems over a clinical examination in discriminating between mild and severe pancreatitis. We agree with this assessment. Our institution admits more than 300 patients with acute pancreatitis each year, and the triage of these patients is based on clinical examination results. Our experience suggests that prognostic scoring systems do not play a useful role as a discriminating tool between acute and severe pancreatitis at hospital admission.

During the past 2 decades, the outcome of severe acute pancreatitis has significantly improved and the overall mortality has significantly decreased to less than 15% in patients with severe pancreatitis. A small subgroup accounts for most of the deaths, and this includes patients who develop a fulminant course (3 of the 14 deaths in our study) and patients who require multiple debridements associated with prolonged ICU care and hospitalization and a high mortality. To our knowledge, the role of the prognostic scoring systems has not been previously defined in identifying the subgroup with a poor outcome in a predetermined population of patients with severe pancreatitis. With this intent, in the present study, we examined the relationship between the APACHE II score and an adverse outcome in a group of patients with severe pancreatitis. In this study, we used the Atlanta classification of severe acute pancreatitis, and 85% of the patients in this study had severe pancreatitis by these criteria. As a group, our patients were ill: 41% were admitted to the ICU and evidence of organ failure (≥1 organ) developed in 52%. The 48-hour and combined admission and 48-hour APACHE II scores accurately predicted the presence of significant pancreatic necrosis, organ failure, and death. Furthermore, patients dying of acute pancreatitis had significant deterioration in their APACHE II score during the initial 48 hours after admission compared with survivors. Deteriorating APACHE II scores at 48 hours after admission were found in 7 of 9 patients who had late deaths. Our results suggest a role for the APACHE II score in risk stratification in patients with severe pancreatitis. Furthermore, deteriorating APACHE II scores in the first 48 hours after admission may be an ominous sign that may predict late deaths from severe acute pancreatitis.

We propose that a paradigm shift is necessary regarding the utility of prognostic scoring systems in patients with severe acute pancreatitis. Prognostic scoring systems are not beneficial in triaging patients with severe acute pancreatitis. On the other hand, the identification of the subgroup of patients who do poorly will represent a significant leap in the management of severe acute pancreatitis.

The CT severity index has been proposed as an alternate to conventional prognostic scoring systems. A recent study by Simchuck et al¹² has shown the prognostic value of the CT severity index in patients with severe acute pancreatitis. In patients with a CT severity index greater than 5, there was an 8-fold higher risk of death and a 10-fold higher chance of the need for necrosectomy. Because the researchers did not provide sensitivities, specificities, and predictive values of the CT severity index, the value of this scoring system in identifying patients with severe pancreatitis who do poorly is unclear. In our study, while highly significant associations were found with 3 measures of outcome and the overall predictive accuracy was high, the sensitivities and positive predictive values were low. Clearly, improved risk stratification is necessary for patients with severe acute pancreatitis. Future research directed at CT-based morphological criteria, pathophysiological scoring systems, or a combination of both may identify the subgroup of sick patients with an adverse outcome in those with severe acute pancreatitis and may provide an opportunity for targeted therapy to improve care for these patients.

This study has demonstrated the improved predictive value of the APACHE II score at 48 hours after admission compared with the admission score at predicting the presence of pancreatic necrosis, organ failure, and death in a preselected group of patients with severe pancreatitis. Furthermore, a deteriorating APACHE II score was significantly associated with mortality in patients with severe acute pancreatitis. Serial determination of the APACHE II score in patients with severe acute pancreatitis may identify the subgroup with an adverse outcome. Refinement of prognostic indicators for risk stratification in patients with severe pancreatitis may lead to improved care for the subgroup of patients with severe pancreatitis who have a poor outcome.

This paper was presented at the 109th Scientific Session of the Western Surgical Association, San Antonio, Tex, November 13, 2001.

Corresponding author and reprints: Dilip Parekh, MD, Department of Surgery, The University of Southern California, 1510 San Pablo St, Suite 430, Los Angeles, CA 90033 (e-mail: dparekh@surgery.usc.edu).