Coffey, Michael J.*; Nightingale, Scott†; Ooi, Chee Y.‡
See “Predicting the Severity of Pediatric Acute Pancreatitis: Are We There Yet?” by Uc on page 584.
Acute pancreatitis (AP) remains a poorly understood condition in adults and even more so in children. The diagnosis of AP in children can be challenging and requires a high index of suspicion. Presenting symptoms are heterogeneous, nonspecific, and dependent on the patient's age and/or developmental level (1). Although AP is less frequent in children than in adults, an increasing incidence of AP in the pediatric population has been reported (2–7). Because AP can result in significant morbidity and mortality, early prediction of severity is important for management considerations.
Several prognostic systems have been developed and validated for adult AP, the most widely used being the Ranson criteria (8,9), the Glasgow (10) and Modified Glasgow (11), and the Acute Physiology and Chronic Health Evaluation II score (12). Unfortunately, these systems are not always practical for use in children.
DeBanto et al (13) were the first to develop and validate a scoring system to predict the severity of AP in children. Severe outcome was predicted by meeting ≥3 of 8 criteria based on the following parameters: age, weight, admission white blood cell count (WCC), admission lactate dehydrogenase, 48-hour trough Ca2+, 48-hour trough albumin, 48-hour fluid sequestration, and 48-hour rise in serum urea nitrogen. This scoring system is limited by the need for repeated testing over a 48-hour period, and the inability to use it within 48 hours of presentation. Furthermore, age and weight are considered separate criteria, even though they are highly correlated in a pediatric population. This pediatric-specific scoring system, when compared with the adult-based Ranson and Glasgow scores, was reported to have better, although still low, sensitivity (70% vs 30% and 35%, respectively) and negative predictive value (NPV) (91% vs 85% and 85%); however, subsequent comparative studies have found that the performance of the DeBanto et al scoring system was not superior to adult-based scoring systems and may not be adequately accurate for the purposes of risk stratification (14,15).
Our study aimed to identify early (within 24 hours of hospital presentation) predictors of the course of severity in pediatric AP, with validation analysis in a separate cohort of pediatric patients.
A retrospective review (January 2000–July 2011) was performed in all patients admitted to the Sydney Children's Hospital Randwick and John Hunter Children's Hospital, which served as the derivation and validation groups, respectively. Both hospitals are tertiary referral hospitals for their respective regions in the state of New South Wales, Australia. The present study was approved by the human research ethics boards of both participating institutions, South Eastern Sydney Human Research Ethics Committee (10/188) and Hunter New England Human Research Ethics Committee (11/02/16/5.07).
Patients younger than 18 years at time of presentation were eligible for inclusion if they had a diagnosis of AP or recurrent AP (RAP). AP was defined as abdominal pain not due to other causes, plus either elevated serum lipase or amylase ≥3 times the upper limit of the normal reference range (ULN) or imaging evidence of pancreatitis, or both (16). Recurrent AP was defined as ≥2 separate documented episodes of AP, or 1 documented episode with a specified history of a previous AP episode (16). Complete resolution of pain and at least a 1-month pain-free interval between episodes was required to be considered RAP. Each documented episode of RAP was analyzed as a separate AP episode. Patients presenting with pain and elevation of serum pancreatic enzyme levels because of pseudocyst(s) were excluded.
Demographic, clinical, laboratory, and radiographic data were collected from the medical records of patients with a confirmed diagnosis of AP. Laboratory data within 24 hours of initial hospital presentation were analyzed, including data from referring hospitals if presentation occurred outside the study institutions. Unavailable data for a given parameter were recorded as missing.
Determination of Severity
An episode of AP was defined as severe if the patient developed local complications such as necrosis, infected necrosis, hemorrhage, or abscess formation evident on imaging; required pancreatic surgery; died because of complications of pancreatitis; was admitted to an intensive care unit; or developed organ dysfunction (systolic blood pressure <90 mmHg or evidence of pulmonary insufficiency [PaO2 <60 mmHg, or oxygen requirement with chest radiographic abnormalities, eg, atelectasis and/or pleural effusion]) (17). An episode was considered severe only if the intensive care unit admission and/or organ dysfunction primarily resulted from pancreatitis instead of comorbid conditions.
To compare mild and severe episodes of AP, continuous variables were evaluated using unpaired Student t tests or Mann-Whitney U tests depending on the normality of data distribution. Because of different reference ranges for serum lipase and amylase at each hospital laboratory, values for lipase and amylase were presented as the ratio above the ULN. Each categorical variable was examined using a Fisher exact test. A P value <0.05 was considered statistically significant; however, Dunn-Sidak correction was applied to account for multiple comparisons. Parameters that had a statistically significant correlation with severe disease were analyzed using receiver-operating characteristic (ROC) curves to determine optimal cutoffs based on the area under the ROC. Cutoffs for univariate relations were then examined using binary logistic regression analysis to determine significance (P < 0.05) and calculate odds ratios (OR) with 95% confidence intervals (CI). Any parameter with >25% missing data was excluded and all significant parameters were tested for collinearity (with a variance inflation factor >2 considered problematic) before inclusion in a multivariate binary logistic regression model. An OR with 95% CI, sensitivity, specificity, positive predictive value (PPV), NPV, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) was calculated for any significant parameter within the model. All of the statistical calculations were performed in SPSS 19.0 (SPSS Inc, Chicago, IL).
A total of 73 AP episodes from 69 patients were identified from the derivation institution (Sydney Children's Hospital). Twenty-five of the 73 (34%) AP episodes met the criteria for severe AP. The demographic data for the mild and severe outcomes, along with the distribution of etiologies, are summarized in Table 1. The distribution of criteria for classification as severe disease is presented in Table 2.
For the validation sample (John Hunter Children's Hospital), there were 58 episodes of AP from 56 patients during the same time period. Fourteen of the 58 (24%) AP episodes met the criteria for severe AP. The characteristics of the validation cohort are summarized in Tables 1 and 2.
Derivation and Validation Cohorts
There were 131 episodes of AP from 125 patients in the derivation and validation cohorts; 29.8% of these episodes met the criteria for severe AP. The median age (interquartile range) of all pancreatitis episodes was 11.6 (8.0–13.7) and 15.1 (11.2–17.2) for the derivation and validation cohorts, respectively (P < 0.001). The sex distribution of boys (% males) for the derivation and validation cohorts was 46 (63%) and 23 (40%), respectively (P = 0.009).
Derivation of Predictor(s) of Severe AP
Comparisons between mild and severe episodes of AP within the derivation cohort were performed for all of the demographic and laboratory parameters (Table 3). To account for the multiple comparisons, a Dunn-Sidak correction was applied to the criterion α cutoff of 0.05, lowering it to 0.0017. Only serum lipase was found to be statistically significant after Dunn-Sidak correction when comparing mild and severe episodes of AP (Table 3); however, 5 other variables with low P values (<0.05) and biologic plausibility (serum amylase, phosphate, WCC, neutrophils, and hemoglobin) were also of interest. Using ROC curve analysis, a cutoff was determined for each of these parameters (Table 4). Because statistically significant cutoffs for WCC and neutrophils could not be determined on ROC curve analysis, these were excluded from further analysis.
Based on these cutoffs, univariate logistic regression analysis was used to assess each parameter (Table 5). All parameters with their defined cutoffs were statistically significant.
Several parameters were excluded from the final multivariate regression model. Serum phosphate was excluded because there were >40% missing data. Tests for collinearity revealed a relation (variance inflation factor ≥2) between lipase and amylase. Because lipase appeared to be a better predictor of severity on area under the ROC analysis, amylase was excluded from further analysis. The final 2 remaining parameters, lipase and hemoglobin, were analyzed using multivariate logistic regression (Table 5). Only serum lipase ≥7 × ULN was associated with a statistically significant increased risk of severe AP (P = 0.006), with an OR (95% CI) of 10.1 (2.0–51.6) within the multivariate model.
Validation of Predictor(s) of Severe AP
Multivariate analysis was repeated in the validation cohort using the same parameters of serum lipase and hemoglobin. Serum lipase was found to be a statistically significant risk factor for severe AP (P = 0.025), with an OR (95% CI) of 7.3 (1.3–41.5) (Table 5). Hemoglobin was also significant (P = 0.041), with an OR (95% CI) of 0.1 (0.01–0.9).
Serum Lipase ≥7 × ULN as a Predictor of Severe AP
Serum lipase ≥7 × ULN (within 24 hours from time of presentation) was the only statistically significant variable after Dunn-Sidak correction and the only significant predictor in both derivation and validation cohorts. Thus, the performance and clinical applicability of serum lipase as a predictor of course of severity in pediatric AP (ie, sensitivity, specificity, PPV, NPV, PLR, and NLR) were determined. Based on the combined derivation and validation data, serum lipase ≥7 × ULN (within 24 hours from time of presentation) was associated with an OR of 7.1 (95% CI 2.5–20.5; P < 0.001) for developing severe AP. Sensitivity, specificity, PPV, NPV, PLR, and NLR were 85%, 56%, 46%, 89%, 1.93, and 0.27, respectively. Of the 131 patients included in this combined cohorts analysis of serum lipase, 23 patients (17 mild and 6 severe) did not have a lipase recorded within 24 hours of presentation and were treated as missing data.
In the present study, we report the potential utility of serum lipase as an early predictor of severity of pediatric AP. We observed that serum lipase, performed within 24 hours of presentation, was significantly associated with severity of AP in children and adolescents. Those with serum lipase concentrations ≥7 × ULN were significantly more likely to follow a severe course of AP (OR 7.1, P < 0.001). This finding was derived as an independent predictor of severity of AP using multivariate regression and confirmed in a contemporaneous cohort from a separate pediatric hospital. The sensitivity of 85% and NPV of 89% for serum lipase concentrations ≥7 × ULN in predicting severe AP suggest that serum lipase may be used as a screening tool to rule out most severe cases of AP in children. Because the true prevalence of AP in the pediatric population is unknown and may vary geographically (2,4,18), we also presented likelihood ratios (which are independent of disease prevalence) to allow application to different populations and settings. In our study population, the NLR of 0.27 suggests a negative posttest probability of 10.6% (ie, 89.4% of patients with a serum lipase result ≥7 × ULN will have a severe AP course).
The present study is clinically relevant because it may provide a simple tool for screening pediatric patients who are unlikely to develop severe AP. Serum lipase is routinely ordered for the diagnosis of pancreatitis, and therefore its benefit as a prognostic tool is without any additional cost. The ability of serum lipase levels to predict AP severity within 24 hours of presentation can provide clinicians with prognostic information early in the course of disease. This is in contrast to many other scoring systems that require data to be collected at later time points. Predicting severity of AP early in the course of disease is critical to optimizing therapy and deciding intensity of monitoring required. Although the management of AP is primarily supportive, recent reports in adult AP demonstrated improved outcomes for severe AP with early interventions such as early aggressive intravenous fluid resuscitation (19) and early enteral feeding (within 1–2 days after developing severe AP) (20–22).
Hemoglobin was not statistically significant on univariate and multivariate analyses of the derivation cohort; however, it was statistically significant in the smaller validation cohort (OR 0.1, P = 0.041). We speculate that this is because of the increase or decrease in hemoglobin, which may occur in patients with severe AP. Increases may occur with intravascular hypovolemia (from 3rd space losses and/or dehydration), whereas decreases may occur with hemorrhagic pancreatitis, gastrointestinal hemorrhage, or hemodilution from aggressive intravenous fluid resuscitation.
In the largest comparison of scoring systems based on pediatric patients with AP, the DeBanto, Ranson, and Modified Glasgow scoring systems performed with: sensitivities of 48.2%, 51.8%, and 51.8%, respectively; specificities of 77.4%, 86.5%, and 88.4%, respectively; PPVs of 43.5%, 58.0%, and 61.7%, respectively; and NPVs of 80.5%, 83.2%, and 83.5%, respectively for severe AP (14). Similar findings were also observed in another separate but smaller comparative study based on Japanese children with AP (15). Hence, the performance of the pediatric-specific DeBanto scoring system was not superior to adult-derived scoring systems as initially reported. Therefore, serum lipase ≥7 × ULN has the advantage over other scoring systems in pediatric AP because it provides a simple, highly sensitive tool, with high NPV, to rule out most cases of severe AP within the first 24 hours of presentation. The NPV of 89% suggests an 89% probability of not developing severe AP in a patient with serum lipase <7 × ULN; however, the low specificity and PPV means that a serum lipase level ≥7 × ULN does not imply a high likelihood of developing severe AP; indeed, only 46% of such patients in our cohort had severe AP. Nonetheless, serum lipase may be used to complement other existing predictive systems because of the differences in performance (sensitivity and specificity) and optimal timing of application. Serum lipase may be used for risk stratification (to rule out most cases of severe AP) within 24 hours of presentation, whereas scoring systems such as the DeBanto, Ranson, or Modified Glasgow could be applied after 48 hours to provide adjunctive prognostic information.
Of note, serum lipase and amylase have not been shown to correlate with severity of AP in previous adult studies (23,24); however, the differences in study design and methodology may explain this discrepancy. Lankisch et al (24) arbitrarily used ≥3 × ULN as cutoffs for both serum lipase and amylase to predict severity, whereas Winslet et al (23) evaluated serum amylase rather than serum lipase levels. In our study, we found serum lipase to be a better predictor of severity than amylase within 24 hours from time of presentation. One explanation for this is that the half-life of lipase (6.9–13.7 hours) is much longer than amylase (2–2.2 hours) (25), making fluctuations less likely. A recent pediatric study also reported lipase to be a better diagnostic marker of pancreatitis than amylase, particularly in infants (26). A positive correlation between serum pancreatic enzyme levels and disease severity has also been observed in a murine study, in which taurocholate-induced severe necrotizing pancreatitis was compared with cerulein-induced mild pancreatitis and a sham control (27). Both serum lipase and amylase were significantly higher in the taurocholate group than the cerulean-induced group at 6, 12, and 24 hours postadministration. Interestingly, this difference was not seen at 48 hours because serum enzyme levels had returned to normal. These findings provide support for a correlation between severe forms of pancreatitis and higher serum pancreatic enzyme levels, as well as the assessment of serum pancreatic enzyme levels early in the course of disease. It may also explain why previous studies, which used a later cutoff (eg, 48 hours), did not find this association. The importance of lipase measurement in clinical practice is also highlighted by 2 recent studies demonstrating lipase elevation to be predictive of biliary tract disease in both adults (28) and children (29).
The proportion of severe cases of AP in our combined cohorts was 29.8%, marginally higher than previous reports of 20% to 28% (13–15). We suggest that this may be because of the fact that the derivation and validation cohorts were derived from tertiary pediatric referral hospitals. Although there is presently no consensus on the definition of severe AP in children, the criteria used in the present study were similar with previous studies (13,14). Our criteria were designed to focus on the development of local (excluding pseudocyst formation) and systemic complications, and the need for pancreatic surgery (distal pancreatectomy and laparotomy). Consensus guidelines in pediatric AP are required to improve the quality and reproducibility of future large, multicenter, prospective studies.
The present study has several limitations. Because it is retrospective, not all patients had blood tests performed within 24 hours of presentation and data collection was limited by the availability of medical records. There were 8 children (6 severe and 2 mild AP cases) who were referred to the study institutions after presenting initially at a referring center. Results for these patients were taken from the referring institution to ensure results were within 24 hours of initial presentation. As previously mentioned, 23 of 131 patients (17.5%) did not have a serum lipase recorded within 24 hours of presentation. Because of the small sample size, other potentially significant parameters may not have reached statistical significance. Nonetheless, we collected data from 2 separate major pediatric referral centers to test validity and ensure generalizability of our findings. Further large multicenter prospective studies involving subjects from different demographics and regions are needed to properly validate our findings, as well as to determine other potential markers to predict the severity of pediatric AP. In the present study, we did not investigate the effect of etiology on severity of AP because our aim was to identify early predictors of severity based on laboratory parameters available in the first 24 hours of presentation. Moreover, the etiology of pancreatitis is often unclear in children, especially early in the course of the disease. With larger multicenter studies, it may be possible to risk-stratify patients as per etiology. The frequency of common causes of AP in the present study is consistent with previous reports in pediatric pancreatitis (13,30).
In conclusion, we have identified and validated the potential use of serum lipase as a simple early predictor of severity in pediatric AP; those with serum lipase levels <7 × ULN within 24 hours of presentation are highly likely to follow a milder course. The ability to risk-stratify pediatric patients within 24 hours of presentation may allow for early targeted interventions and lead to improved outcomes in pediatric patients with AP.
The authors thank Professor Andrea Rita Horvath and Keith Westbury, South Eastern Area Laboratory Services; Marcin Pasternak, Jacqueline Chao, and Filip Tota, Prince of Wales Medical Records; and Dr Huy Tran, Hunter Area Pathology Service.
Renkin's milk antirickets advertisement, 1925. Photo courtesy of the subject, Philip R. Harris, PhD.
Cod liver oil had been recommended for rickets as early as 1824, but causative factors and cures eluded the medical community until the 20th century. The near absence of rickets from sunny countries, such as Egypt, and fish-consuming countries, such as Japan, made little impression. In 1909, Georg Schmorl (1861–1932) in Denmark examined the bones of infants dying between the ages of 2 and 48 months. He ascertained, to the astonishment of all, that nearly 90% of those studied had evidence of rickets. A series of monographs published between 1908 and 1912 by Russian Isidore Schabad (1870) established that cod liver oil prevented rickets as well as cured it. In Berlin, Kurt Huldschinsky (1883) in 1919 used artificial heliotherapy to treat rickets, and in New York, the research of Alfred Hess (1875–1933) confirmed Huldschinsky's findings.
Biochemical studies continued with the research of John Howland (1873–1926) and Benjamin Kramer (1887–1972) who developed a semimicro method for serum calcium and phosphorus measurements. With the advent of the x-ray, they were able to study the evolution and progress of the osteopathy. Edward Parks (1877–1969) and Elmer McCollum (1879–1967) further refined the work of Howland and Kramer using rats as animal models and studying dietary as well as rachitic factors.
At the 1921 American Pediatric Society meeting, sunlight and cod liver oil antirachitic factor were discussed, triggering a cascade of research into the solar irradiation of dermal provitamins. Vitamin D finally was identified as the curative agent, found in the barely palatable cod liver oil. When Philip C. Jeans (1883–1952) in 1936 convinced the Commission on Foods and the American Medical Association to mandate supplemental vitamin D be added to cow's milk, future generations of children were spared the disagreeable taste of cod liver oil and the development of rickets.
—Contributed by Angel Colón, MD
1. Kandula L, Lowe ME. Etiology and outcome of acute pancreatitis in infants and toddlers. J Pediatr 2008; 152:106–110.
2. Lopez M. The changing incidence of acute pancreatitis in children: a single-institution perspective. J Pediatr 2002; 140:622–624.
3. Werlin SL, Kugathasan S, Frautschy BC. Pancreatitis in children. J Pediatr Gastroenterol Nutr 2003; 37:591–595.
4. Nydegger A, Heine RG, Ranuh R, et al. Changing incidence of acute pancreatitis: 10-year experience at the Royal Children's Hospital, Melbourne. J Gastroenterol Hepatol 2007; 22:1313–1316.
5. Lowe ME, Greer JB. Pancreatitis in children and adolescents. Curr Gastroenterol Rep 2008; 10:128–135.
6. Park A, Latif SU, Shah AU, et al. Changing referral trends of acute pancreatitis in children: a 12-year single-center analysis. J Pediatr Gastroenterol Nutr 2009; 49:316–322.
7. Morinville VD, Barmada M, Lowe ME. Increasing incidence of acute pancreatitis at an American Pediatric Tertiary Care Center. Pancreas 2010; 39:5–8.
8. Ranson JHC, Rifkind KM, Roses DF, et al. Prognostic signs and the role of operative management in acute pancreatitis. Surg Gynecol Obstet 1974; 139:69–81.
9. Ranson JHC. Etiological and prognostic factors in human acute pancreatitis: a review. Am J Gastroenterol 1982; 77:633–638.
10. Imrie CW, Benjamin IS, Ferguson JC, et al. A single-centre double-blind trial of Trasylol therapy in primary acute pancreatitis. Br J Surg 1978; 65:337–341.
11. Blamey SL, Imrie CW, O’Neill J, et al. Prognostic factors in acute pancreatitis. Gut 1984; 25:1340–1346.
12. Knaus W, Draper E, Wagner D, et al. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13:818–829.
13. DeBanto JR, Goday PS, Pedroso MR, et al. Acute pancreatitis in children. Am J Gastroenterol 2002; 97:1726–1731.
14. Lautz TB, Chin AC, Radhakrishnan J, et al. Acute pancreatitis in children: spectrum of disease and predictors of severity. J Pediatr Surg 2011; 46:1144–1149.
15. Suzuki M, Fujii T, Takahiro K, et al. Scoring system for the severity of acute pancreatitis in children. Pancreas 2008; 37:222–223.
16. Morinville VD, Husain SZ, Bai H, et al. Definitions of pediatric pancreatitis and survey of current clinical practices. J Pediatr Gastroenterol Nutr 2012; 55:261–265.
17. Bradley ELI. A clinically based classification system for acute pancreatitis: summary of the Atlanta International Symposium. Arch Surg 1993; 128:586–590.
18. Morinville VD, Barmada MM, Lowe ME. Increasing incidence of acute pancreatitis at an American pediatric tertiary care center: is greater awareness among physicians responsible? Pancreas 2010; 39:5–8.
19. Warndorf MG, Kurtzman JT, Bartel MJ, et al. Early fluid resuscitation reduces morbidity among patients with acute pancreatitis. Clin Gastroenterol Hepatol 2011; 9:705–709.
20. Wu XM, Ji KQ, Wang HY, et al. Total enteral nutrition in prevention of pancreatic necrotic infection in severe acute pancreatitis. Pancreas 2010; 39:248–251.
21. Petrov MS, Kukosh MV, Emelyanov NV. A randomized controlled trial of enteral versus parenteral feeding in patients with predicted severe acute pancreatitis shows a significant reduction in mortality and in infected pancreatic complications with total enteral nutrition. Dig Surg 2006; 23:336–344.
22. Gupta R, Patel K, Calder PC, et al. A randomised clinical trial to assess the effect of total enteral and total parenteral nutritional support on metabolic, inflammatory and oxidative markers in patients with predicted severe acute pancreatitis (APACHE II > or =6). Pancreatology 2003; 3:406–413.
23. Winslet M, Hall C, London NJM, et al. Relation of diagnostic serum amylase levels to aetiology and severity of acute pancreatitis. Gut 1992; 33:982–986.
24. Lankisch PG, Burchard-Recker IS, Lehnick D. Underestimation of acute pancreatitis: patients with only a small increase in amylase/lipase levels can also have or develop severe acute pancreatitis. Gut 1999; 44:542–544.
25. Frank B, Gottlieb K. Amylase normal, lipase elevated: is it pancreatitis? Am J Gastroenterol 1999; 94:463–469.
26. Park AJ, Latif SU, Ahmad MU, et al. A comparison of presentation and management trends in acute pancreatitis between infants/toddlers and older children. J Pediatr Gastroenterol Nutr 2010; 51:167–170.
27. Wittel UA, Wiech T, Chakraborty S, et al. Taurocholate-induced pancreatitis. A model of severe necrotizing pancreatitis in mice. Pancreas 2008; 36:e9–e21.
28. Cornett DD, Spier BJ, Eggert AA, et al. The causes and outcomes of acute pancreatitis associated with serum lipase >10,000 U/L. Dig Dis Sci 2011; 56:3376–3381.
29. Ma MH, Bai HX, Park AJ, et al. Risk factors associated with biliary pancreatitis in children. J Pediatr Gastroenterol Nutr 2011; 54:651–656.
30. Bai HX, Lowe ME, Husain SZ. What have we learned about acute pancreatitis in children? J Pediatr Gastroenterol Nutr 2011; 52:262–270.