Acute peritonitis (AP) is a major factor for non-traumatic mortality^[1] and one of the most common causes of acute abdomen.^[2]AP is a result of a complicated intra-abdominal infection^[3] and is still associated with high morbidity, mortality, and healthcare costs worldwide^[4]. Based on the spread of infection, it is classified as local or diffuse.^[5] Local peritonitis (LP) may manifest as peritoneal inflammation encapsulated by fibrous tissue containing leucocytes, bacteria, debris, and exudate (abscess), or as a non-encapsulated process involving no more than one intraperitoneal area.^[3]

Globally, mortality rate of AP varies between 10% and 30%.^[1,6,7] This data refers mainly to patients with diffuse peritonitis, while no exhaustive study on the death rate of LP has yet been conducted. Unfortunately, nowadays it is also unclear which might be the prognostic factors of unfavorable outcome in LP. Various prognostic scoring systems have been developed over the years; unfortunately, none of them is widely accepted in everyday practice. No study so far (to the best of our knowledge) has been conducted investigating mortality prediction scores in LP exclusively. Therefore, we set out to explore the prognostic performance of four of the easiest for calculation scoring systems: two peritonitis-specific scores – the Mannheim Peritonitis Index (MPI) and World Society of Emergency Surgery Sepsis Severity Score (WSES SSS), and two disease-independent ones – systemic inflammatory response syndrome (SIRS) and quick sequential organ failure assessment (qSOFA) score in patients with LP.

The Mannheim Peritonitis Index (MPI), developed by Wacha and Linder^[8] in 1983, seems to be one of the oldest and most practical score for patients with secondary peritonitis^[8,9]. The World Society of Emergency Surgery Sepsis Severity Score (WSES SSS) was designed by the aforementioned surgical society in 2014 as a prognostic scoring system specific for cIAIs.^[1] In 1991, the Systemic Inflammatory Response Syndrome (SIRS) was first introduced as criteria of defining sepsis and predicting in-hospital death.^[10] In 2016, a working group created the current definitions of Sepsis-3 and removed the term SIRS from the definition of sepsis.^[11] The same group introduced the quick sequential organ failure assessment (qSOFA) score as a prognostic score that could immediately determine which patients with suspected infection are likely to need intensive care or die in the hospital.^[11]

Thus, in our study we aimed to find out if MPI, WSES SSS, SIRS, and qSOFA could prognosticate a fatal outcome in patients with LP.

Despite the evolution of the diagnostic and management techniques, AP remains a great challenge to emergency surgeons and critical care physicians. It is responsible for nearly 20% of all sepsis cases in Intensive Care Units and is the second most common cause of infectious morbidity and mortality after pneumonia.^[4] An early prediction of mortality allows doctors to identify those patients with AP that are more likely to die during hospitalization and to change the inadequate management strategy so that a fatal outcome may be avoided. In Europe, LP occurs in 63.5% of patients with AP^[13], and the international studies report the range between 56.4% and 64%^[1,6,12]. Although approximately 2/3 of patients with complicated intra-abdominal infections (cIAIs) have LP, we could not find any study that analyzes prognostic factors or scores in patients with LP exclusively. We chose to assess the predictive ability of four scoring systems which are simple and very easy to calculate. The MPI, introduced by Wacha and Linder^[8], represents an independent, objective, and effective score for predicting mortality, which has shown superiority over other scoring systems in AP.^[8,9] The WSES SSS, developed in 2014, was already validated in several studies^[12,14] and was considered a precise and practical prognostic score for cIAIs. The SIRS was designed to define sepsis and predict mortality.^[10] In 2016, the Sepsis-3 redefinition task force removed SIRS from this definition and introduced qSOFA as a rapid score that could almost instantly determine the need for intensive care or the risk of in-hospital death.^[11]

We observed the qSOFA score as not helpful prognostic tool in LP. The ROC Curve Analysis revealed a very low predictive value (AUROC=0.571), whereat only one of non-survivors had qSOFA ≥2 points (16.7%). No significant differences were found between survivors and non-survivors according the qSOFA values (p=0.355). We found no research that studies prognostic performance of this score in LP. However, in patients with cIAIs, Tolonen et al.^[15], Jung et al.^[16], and Raimondo et al.^[17] observed a better predictive value of the qSOFA score: AUROC=0.723, AUROC=0.717, and AUROC=0.722, respectively.

Similar findings were established for the predictive ability of SIRS (AUROC=0.483), and SIRS ≥2 was observed both in 1/3 of survivors and non-survivors (32.3% vs. 33.3%, p=1.00). We found no data about prognostic performance of SIRS in LP in the available literature. Although SIRS was not developed as a prognostic scale but as a tool for defining sepsis, over the years it has been studied as a predictor of death in different clinical settings. In patients with cIAIs, Jung et at.^[16] and Raimondo et al.^[17] reported higher value of the AUROC Curves with a poor ability to prognosticate mortality: AUROC=0.672 and AUROC=0.692, respectively.

A fair prognostic accuracy was demonstrated in the present study by WSES SSS (AUROC=0.734). Its optimal cut-off value was WSES SSS ≥4 points and it permitted prediction of adverse outcome with a sensitivity of 83.3% and a specificity of 62.9%. No study (to our knowledge) explores the predictive performance of WSES SSS in LP yet. In patients with cIAIs, several authors reported a better accuracy: Godinez-Vidal et al.^[14] – AUROC=0.931 with a sensitivity of 76.47%, a specificity of 90.48%, Raimondo et al.^[17] – AUROC=0.887, a sensitivity of 85.7% and a specificity of 75.9%, and Tolonen et al.^[15] – AUROC=0.809, a sensitivity of 73% and a specificity of 76%. Godinez-Vidal et al.^[14] and Raimondo et al.^[17] reported the same threshold as ours, while Tolonen et al.^[15] found a much higher threshold (≥8). The median WSES SSS in the present study was higher in non-survivors compared to survivors [6 (4.25-8) vs. 3(0-6)], and the difference was very close to significance (p=0.054). We suggest that this could be due to the small number of surveyed patients.

Among the four scores, MPI showed the best ability to prognosticate the fatal outcome in LP (AUROC=0.805) with a sensitivity and a specificity of 66.7% and 80.6%, respectively. Furthermore, the mean score in non-survivors was significantly higher than those in survivors (25.33±4.97 vs. 17.98±4.79, p=0.012). We found no other study that investigated the prognostic value of MPI in LP. Budzyński et al.^[9] observed in patients with secondary peritonitis, a predictive accuracy similar to ours (AUROC=0.810). A better prognostic values were reported by Salamone et al.^[18] in AP – AUROC=0.89 and Godinez-Vidal et al.^[9] in cIAIs – AUROC=0.843, while Tolonen et al.^[15] reported a lower value in severe cIAIs – AUROC=0.774. In the original study of Wacha and Linder^[8] the determined cut-off value was MPI=26 points with a sensitivity of 84% and a specificity of 79%. We identified the same threshold with a sensitivity of 66.7% and a specificity of 80.6%. Lower threshold was reported by Godinez-Vidal et al.^[14] – MPI ≥18 points with a sensitivity of 82.35% and a specificity of 79.17%, and Salamone et al.^[18] – MPI=20 with a sensitivity of 78% and a specificity of 89%. Higher cut-off values were reported in the studies of Tolonen et al.^[15] – MPI ≥30 with a sensitivity of 51% and a specificity of 79%, and Budzyński et al.^[9] – MPI=32 with a sensitivity of 66.7% and a specificity of 97.9%.

The observed in-hospital mortality rate in our study was 8.8%. Pupelis et al.^[19] reported a little bit higher value than ours – 9.4% in patients with LP. Maseda et al.^[20] observed a death rate of 11.1% in critically ill patients with LP. The highest mortality rate was reported by Blot et al.^[7] in critically ill patients with LP – 24.2%. Unfortunately, none of these studies showed other data about the predictive performance of scoring systems in LP.

ROC Curve analysis in the present study pointed prognostic superiority of MPI to WSES SSS, qSOFA, and SIRS (AUROC=0.805 vs. 0.734 vs. 0.571 vs. 0.483), whereat it is the only score with good ability to discriminate non-survivors (AUROC of MPI is greater than 0.8). The performed bivariate correlation analysis showed one significant correlation – between outcome and MPI (p=0.012, r=0.302), and the others were weak or very weak with no significance.

This is the first study (to the best of our knowledge) which analyzes prognostic performance of MPI, SIRS, WSES SSS and qSOFA and investigates the correlations between outcome and these scores in patients with LP.

As limitations of our study we can highlight the small number of investigated patients, the single-center experience, and the retrospective design

In patients with LP due to cIAIs, WSES SSS, SIRS and qSOFA score show no ability to predict the adverse outcome. Although MPI is the oldest among surveyed scores, it shows the best ability to recognize patients at higher risk of death.