REPLY

Article in PDF format - JOP Home page

Reply to: Assessment of the Severity of Acute Pancreatitis. The Usefulness of ROC Analysis in Comparative Studies of Clinical and Imaging Prognostic Indices

Gunay Gurleyik¹, Gamze Kilicoglu²

¹Department of Surgery and ²Department of Radiology, Haydarpasa Numune Education and Research Hospital. Istanbul, Turkey

Dear Sir:

We would like to thank Brestas et al. for their interest in our article published in JOP. J Pancreas (Online) [1]. In this study, The results of an imaging method (computed tomography), a clinical scoring system (APACHE II) and a biochemical measurement (serum CRP concentration) were evaluated in a group of patients with acute pancreatitis. The accuracy of these variables for predicting the clinical course of the disease was calculated. This had initially been classified using the Atlanta criteria. Brestas et al. presented their opinions and suggestions mainly on two subjects:

• the cut-off values of the computed tomography severity index (CTSI) and APACHE II;

• evaluation of the results using the receiver operating characteristics (ROC) curve.

We chose these cut-off values because of our personal experience [2] on acute pancreatitis cases (length of hospital stay, morbidity and mortality, etc.) and the results of previous studies. Larwin and McMahon [3] found that patients with an APACHE II score greater than 7 were likely to have a severe clinical course while Connor et al. [4] reported increasing mortality in patients with APACHE II scores greater than 7. Balthazar et al. [5] reported that acute pancreatitis cases with a CTSI of 0 to 2 points had 2% morbidity and no mortality while morbidity and mortality rates for patients with a CTSI ranging from 0 to 3 were 8% and 3%, those with a CTSI of 4-6 were 35% and 6% and those with a CTSI of 7-10 were 92% and 17%, respectively. Therefore, aiming to have a lower mortality rate for our patients, we preferred to use a CTSI score of 3 and an APACHE II value of 7 as our cut-off points.

The variables of our study were compared by valid statistical methods. The sensitivity, specificity, positive and negative predictive values and accuracy were also calculated for each variable. We agree with Brestas et al. that the ROC curve could have been used, especially in assessment of the imaging technique. Thus, we applied the test to our results.

The area under the ROC curve (AUC±SE evaluated by SPSS 10.0) for CTSI was 0.969±0.024, clearly supportive of the high accuracy of this index in predicting the severity of acute pancreatitis (Figure 1). A value of the CTSI equal to 3 was identified as the best cut-off using the procedure proposed by Pezzilli et al. [6] (the maximum likelihood ratio, LR, was 10.3). This value corresponds to previously published values of sensitivity and specificity [1]. The APACHE II gave an AUC value of 0.812±0.074 and a best cut-off value of 7 (the maximum LR was 3.6). Sensitivity, specificity, and frequency of cases correctly identified by applying this cut-off are 61.5% (8 out of 13 severe acute pancreatitis patients), 95.2% (40 out of 42 mild pancreatitis patients), and 87.3% (48 out of 55 overall acute pancreatitis patients), respectively.

Figure 1. Receiver operating characteristics (ROC) curves of the computed tomography severity index (CTSI) and APACHE II in distinguishing between mild and severe acute pancreatitis. Data from Gurleyik et al. [1]. Red bullets show the best cut-off values. Green bullets shows the point of the ROC curve corresponding to the values of previously published APACHE II score [1].

Finally, it should be noted that the AUC of the CTSI was significantly higher when compared with the APACHE II score (P=0.044, z-test).

In conclusion, the use of ROC curve analysis confirmed our previous results [1] showing that values of the CTSI greater than 3 are highly indicative of severe acute pancreatitis. On the other hand, the ROC curve applied to our data indicates values of an APACHE II score greater than 7 are necessary in order to identify severe acute pancreatitis patients instead of values equal to or greater than 7 as we used in our paper [1]. Moreover, both the CTSI and the APACHE II scores are highly accurate in predicting the natural outcome of specific patients with acute pancreatitis, with the indicated cut-off values clearly specific for the purpose. Finally, the CTSI is significantly more accurate than the APACHE II score.

Article in PDF format (Letter + Reply)
Look up who cited this article

Received January 24^th, 2006

Keywords APACHE; Pancreatitis, Acute Necrotizing; ROC Curve; Tomography, X-Ray Computed

Correspondence
Günay Gürleyik
Eski Bağdat cad. 29/9
Altıntepe 34840 Istanbul
Turkey
Phone: +90-216.489.8325
Fax: +90-216.373.1096
E-mail: ggurleyik@yahoo.com