Predictive value of drain pancreatic amylase concentration for postoperative pancreatic fistula on postoperative day 1 after pancreatic resection

Supplemental Digital Content is available in the text


Introduction
Although perioperative management of patients undergoing pancreatoduodenectomy (PD) has been improved, morbidity still ranges from 20% to 50%. [1][2][3][4] Postoperative pancreatic fistula (POPF), which develops in a range from 16% to 28% of patients undergoing PD, remains the major fatal complication. [5][6][7] POPF was classified to grades A, B, and C by the International Study Group on Pancreatic Fistula (ISGPF). [8] Grade A POPF is biochemical fistula, which does not have any adverse consequences. Grade B and C fistulas are generally designated as clinically relevant POPF (CR-POPF), which usually demand percutaneous drainage and hardly, laparotomy. [8] So far, it has been proved that early drain removal (postoperative day (POD) <4) decreases incidence of complications, compared with late drain removal (POD >5). [9] Usually, drains will be removed according to surgeon's discretion after excluding the risk of POPF. The prognosis of POPF can be advantageous to manage drains removal, enhance recovery pathway, and promote hospital discharge. [9,10] Recently, many studies show high interests in drain pancreatic amylase concentration on POD 1 (DPA1) for the prediction of POPF. Although DPA1 has been implied with superb specificity and sensitivity for overall POPF and CR-POPF, there is still controversy in inconsistent opinions. The present meta-analysis especially aims to assess the value of DPA1 to predict POPF after PD.

Methods
The PRISMA statement and appropriate methods for metaanalysis were followed. [11,12] The ethical statement is not necessary for this meta-analysis.

Inclusion and exclusion criteria
Studies were selected by inclusion criteria as follows: prediction of DPA1 for POPF after PD; POPF recorded and defined as grade A, B, and C according to ISGPF; articles published in English language in peer-reviewed journals. Editorials, case reports, expert opinions, letters, abstracts, and studies without sufficient data to assess predictive value of DPA1 were excluded.

Quality assessment
The QUADAS criteria was accorded to evaluate qualities of involved studies. [13]

Data collection and statistical analysis
Data including sensitivity, specificity, and cutoff values of DPA1 for the prediction of POPF or CR-POPF were documented. STATA 12.0 was used for statistical analysis. The following figures were calculated: sensitivity, specificity, positive likelihood ratio (LR), negative LR (with corresponding 95% confidence interval), pretest probabilities, corresponding posttest probabilities, Cochran Q test, inconsistency index (I 2 ), and area under the receiver operating characteristic curve (AUROC). Deeks funnel plot asymmetry test was performed to assess publication bias.
Two investigators (Y Liu and Y Li) independently extracted the data, and disagreements were settled by discussion with each other.

Publication bias
Deeks funnel plot asymmetry test indicates there is no publication bias among the studies in Fig. 5. Table 3 Meta-analysis of predictive data for overall POPF and CR-POPF.

Discussion
POPF is still a potentially fatal complication, which may increase financial utilization after pancreatic resection. There already arise controversies about intraperitoneal drains following PD. In a recent study, [29] it indicated that early drains removal (POD 4) had significant benefits on decreasing incidence of POPF. Furthermore, in a prospective study with 84 patients who were performed PD, it revealed that it is safe to pull out drains on POD 3 following PD with a lower incidence of POPF in patients with DPA1 5000 U/L. [19] Early prediction of POPF can significantly benefit the patient following PD; however, few studies have assessed the predictive accuracy of DPA for developing of POPF.
Several markers, such as DPA, CRP, WBC, have been proposed as predictors for POPF. [5][6][7]28,30,31] It was implied by Molinari et al [23] that DPA1 > 5000 U/L had a respectively sensitivity and specificity of 93% and 84% for the prediction of POPF following PD. Besides, Ansorge et al [5] recommended serum CRP association with DPA to predict CR-POPF. With the comprehensive consideration, this meta-analysis aimed to assess the accuracy of DPA1 for the prediction of POPF. Up to now, there are few studies to assess the pooled performance of DAP1 for overall and clinically relevant POPF.
In this meta-analysis, DPA1 displayed an outstanding capability in identifying POPF with a high positive LR, which could be acted as a rule-in means for the diagnosis of POPF. Meanwhile, it also showed an acceptable sensitivity and specificity. When the pretest probability was set at 50%, DPA1 indicated an accurate diagnosis of overall POPF in 81% patients and misdiagnosis only in 16% patients by Fagan plot analysis, besides, it also showed accuracy for CR-POPF in 86% positive patients and misdiagnosis in 26% patients. With comprehensive consideration of the pooled results, DPA1 is an appropriate marker for the prediction of POPF. Certainly, more randomized controlled trials should be implemented to provide evidence.
In the present study, it supplies beneficial information to help researchers and clinicians to predict POPF by DPA1. However, there are several limitations in this meta-analysis. First, few studies are assessed as high quality to offer unbiased data. Second, studies involved in this meta-analysis had a vast range in cut-off values. Besides, few studies provided specific amylase range or cutoff at each grade of POPF. Therefore, it is necessary to further explore the values of DPA1 and other markers for predicting the grade of POPF in randomized studies.
It is concluded that DPA1 is a valuable marker to predict POPF, and more randomized controlled trials should be implemented to provide unbiased evidences.