The incidence of peptic ulcer disease has decreased, particularly in Western countries after the introduction of proton-pump inhibitors (PPI) along with antibiotic therapy for Helicobacter pylori eradication. Nowadays, peptic ulcer disease has a reported annual incidence between 0.03 and 0.19% worldwide [1,2,3,4,5,6]. However, although the number of patients with ulcer is decreased to less than one third, there was no marked decrease in the number of deaths from ulcer, and perforated peptic ulcer (PPU) remain a surgical emergency associated with increased mortality, accounting for 37% of all peptic ulcer-related deaths [7]. This indicates that the clinical picture of ulcer became more severe, in particular, combined use of LDA (low-dose aspirin) and non-steroidal anti-inflammatory drugs (NSAIDs) and advanced age serve as risk factors for the occurrence of LDA-induced ulcer and also increase the risk of haemorrhage and aggravation [8,9,10,11,12]. Surgery remains the standard of care for patients with PPU. The benefits of laparoscopic approach (LA) for patients with perforated peptic ulcer have been well-established even in the elderly [13,14,15,16]. The first laparoscopic repair of PPU was reported by Mouret et al. [17]. Laparoscopy is associated with less intraoperative blood loss, better pulmonary function, reduced postoperative pain, quicker return of bowel function, a shorter hospital stay, and a lower incisional hernia incidence when compared with standard open surgery [18]. Thereafter, retrospective studies found acceptable outcomes of LA for PPU. However, because of inconsistent results with specific regard to some technical aspects of such technique, i.e., peritoneal lavage, surgeons questioned the adoption of laparoscopic approach to perform ulcer repair [19, 20]. This leads the surgeons to choose the type of approach based on their personal experience. The aim of our study was to critically appraise the use of the laparoscopic approach in PPU treatment comparing it with open procedure.

Material and methods

Study settings and protocol

This research originates from a previous well-consolidated experience [21] thus following a similar methodology, a new collaborative research group was founded. The IGo-GIPS (Italian Group for Gastro-Intestinal Postoperative Surveillance) is a large, nationwide network created with the aim to undertake both prospective and/or retrospective studies investigating the perioperative outcomes of specific topics mainly concerning gastrointestinal surgery [22]. Centres were included on a volunteer basis, and neither investigators nor participating hospitals were paid for their collaboration. Clinical decisions, including operative technique, were always based on the criteria of individual centres and staff surgeons. Although procedures were not standardised per a study protocol, it is important to note that they were likely similar among participating hospitals, with some slight technical differences across institutions seldom taken into account because they were judged to not influence the outcome. The investigators were informed about the objectives of the project and asked for complete details about the surgical management of patients following standard methods and collection protocols as already described. Data regarding patients were prospectively collected from the study participating centres from January 2017 to June 2018, while data regarding other patients from July 2018 to December 2021 were retrospectively retrieved from hospital electronic databases. The former prospective study protocol was approved by the Ethics Committee of Sapienza University of Rome and of all the centres while no formal approval was requested for any other retrospective non-interventional study except in case of specific indication deemed by a single centre. However, a signed consent for the treatment and the analysis of data for scientific purposes was obtained from all patients before surgery. This study was conducted in accordance with the Declaration of Helsinki and its later amendments. All parts of the studies and the present manuscript have been checked and presented according to the checklist for Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) [23].

Inclusion and exclusion criteria and collected data confirmation

For the aim of the present study, we initially retrieved records of all patients having > 18 years with ICD-9-CM code ranging from 531.× to 534.× requiring emergency surgery from January 2017 to December 2021. Furthermore, bleeding ulcer, neoplastic perforation, sole endoscopic procedures, and emergency operations during the course of any other elective surgery were discarded. Other exclusion criteria were the following: age < 18 years; location other than stomach or duodenal bulb; previous upper GI surgery; lack of informed consent for the study participation, if requested; patients participating in other randomised or interventional clinical trials. Submissions made by unconfirmed participants, duplicate submissions, records with more than 5% of missing data were also excluded. Although the patient's demographic information was collected, all data were anonymized before analysis even for centre identification.

Patients’ characteristics, preoperative variables and objectives of this study

Data collected included patient demographic characteristics and clinical variables, procedure details, and outcomes. Demographics variables and clinical data included: age, gender, weight, height, body mass index (BMI), Glasgow Coma Scale (GCS), heart rate, systolic blood pressure, medical and surgical history (comorbidities), common preoperative biochemical blood examination (including C-Reactive Protein [CPR], and arterial blood gas analysis). Procedure details included: site and size of the ulcer, type of surgical approach and procedure performed, timing, conversion rate, peritoneal contamination). Comorbidity was recorded if the condition was being medically treated at the time of admission, or if previous treatment for the condition was described in the admission report. The Age-adjusted Charlson Comorbidity Index (age-CACI) was calculated and a score ≥ 6 was used to categorise patients having a severe comorbid condition. Preoperative risk was assessed with anaesthesiologist-assigned American Society of Anaesthesiologists (ASA). Furthermore, Systemic inflammatory response syndrome (SIRS) was evaluated according to the original consensus study (Sepsis-1) [24]. SIRS criteria ≥ 2 met the definition of SIRS. When appropriated, the Frailty profile was investigated by the 5-modified Frailty Index (5-mFI) and the Emergency Surgery Frailty Index (EmSFI). As previously described [21], the assessment of Activities of Daily Living and the Fried Frailty criteria was considered and in order to simplifying the concerning statements, the item “Altered autonomy” should be referred to the assessment of activities of daily living, while the item “Altered mobility” has been used to outline some of the Fried’s frailty criteria. When statistical analysis was performed, 5-mFI ≥ 0.4 score was used for categorising frailty as a binary variable according to current literature [25]. Sepsis was evaluated according to the qSOFA score. The Shock index, the Age-Shock Index, the Boey score, the Mannheim Peritonitis Index (MPI), the PULP Score, and the Jabalpur score were also calculated. Postoperative complications have been reported and categorised according to the Clavien–Dindo classification system by the study leader in each of the participating centres and the Comprehensive Complication Index was also calculated [26, 27]. Furthermore, morbidity was divided into three groups named as follows: C–D 1–2, C–D 3, and C–D 4. Although morbidity and mortality have been considered as the 30-day standard period definition, adverse outcomes have been reported regardless of the time elapsed from the surgical procedure if reasonably related to it and occurred during the hospitalisation following the main emergency procedure. First, the entire study cohort was investigated by analysing the initial approach selected regardless of the procedure performed. Furthermore, only the suture repair was taken into account because it was the most common technique adopted. Patients undergoing PPU repair were divided into two groups named Laparoscopic approach (LapA) and Open approach (OpenA) and clinical-pathological features of patients in the both groups were compared. The laparoscopic group included all patients who underwent an attempted pure laparoscopic procedure. An open conversion was defined as when a procedure was attempted via the minimally invasive approach but required an open incision to be completed. Patients who required conversion to laparotomy were included and analysed on an intention-to-treat basis. Leakage was defined as when bile or gastric content was detected in the drain output or by means of CT scan with oral water-soluble contrast. No routine use of the latter or of the methylene blue test was adopted. As well as stated above, due to the multicentre observational design of the study, there was not a uniform standardised protocol neither for the suture technique performed nor for the suture material used. However, because the simple suture, the suture plus omental flap, and suture with omental patch were equally distributed in both groups, the lack of uniformity in the technique and material used was not considered as a bias.

Statistical analysis

Statistical analysis was carried out using StataCorp 2019 STATA Statistical Software: release 16 (College Station,TX: StataCorp LLC). Dichotomous data and counts were presented in frequencies, whereas continuous data were presented as mean values ± standard deviations (SD) and/or median with 25–75 Interquartile Range (IQR) and minimum–maximum range. Differences between means were compared using the independent sample Student’s T-test or the Mann–Whitney U test when indicated. Fisher’s exact test or χ2 test, with or without Yates correction, were used to compare differences in frequencies. After an initial entire cohort summary description, the patients managed with gastric resection and those with lavage and drainage were excluded and a deeper statistical analysis was performed only on the patients who underwent the simple suture repair. Firstly, the accuracy of the morbidity and mortality prediction of MPI, Boey score, PULP score, and Jabalpur score was evaluated by receiver-operating characteristic (ROC) curve analysis, and the corresponding area under the curve (AUC) values was compared. Univariate analysis was performed in order to assess the Odds Ratio of any factor for suture leak and conversion to open. Odds Ratio was calculated by means of a 2 × 2 Contingency table for binary variables while by means of logistic regression for those continuous. All variables with a p value < 0.10 at univariate analysis were entered into a backward stepwise logistic regression model. If needed, receiver-operating characteristic (ROC) curve analysis was performed to calculate the cut-off value corresponding to the maximum Youden’s index of continuous variables in order to clarify the potential clinical relevance of such Odds Ratio values.

Thereafter, a propensity score matching was carried out. The Italian Version of IBM Corp. Released 2012. IBM SPSS Statistics for Macintosh, Version 21.0. IBM Analytics (Italy, Segrate, Milan) integrated with SPSS R Essentials for R Statistical Software version 2.14.2 (Foundation for Statistical Computing, Vienna, Austria) was used. The model was constructed to eliminate selection bias between groups and was reported according to the recommendations of Lonjon et al. [28]. Variables influencing decision regarding surgical approach and variables with potential influence on outcomes were assigned propensity scores using a bivariate logistic regression model. The final model included the following variables: sex as exact, age, ASA score, Age-Shock Index, and Age-CACI. We matched propensity scores 1:1 with the use of the nearest neighbour methods without replacement using the closest callipers width to achieve the maximum number of cases without statistical differences in confounders. In this instance, the calliper width was set at 0.2. All tests were two-tailed and a p value ≤ 0.05 was considered statistically significant.

Result

Entire cohort

A total of 509 patients fulfilling the inclusion criteria were evaluated (Fig. 1). The overall mean age was 62.9 ± 17.4 (range 18 to 100 years), 273 (53.6%) were male. A laparoscopy approach (LA) was performed in 212 patients (41.6%) while an open approach (OA) was performed in 276 patients (54.2%). The overall conversion rate was 4.1% (21 patients). Four hundred and fifty-three patients (89.0%) underwent simple repair with or without an omental patch, 50 patients (9.8%) underwent gastric resection, while a peritoneal lavage and drainage was performed in 6 (1.2%) patients. Table 1 reports demographics, clinical and operative data stratified by the type of approach.

Fig. 1
figure 1

Flow chart of clinical study design. (PSM propensity score matching, Lap. laparoscopy)

Full size image
Table 1 Demographics characteristics and clinical data
Full size table

Simple repair group data

A total of 453 (89%) patients underwent PPU simple repair. Among these, a LA was adopted in 49% (222/453 patients) defined as “LapA group” with a conversion rate of 7.2% (16/222 patients), while an open approach was adopted in 51% (231/453 patients) defined as “OpenA group”.

Table 2 summarises demographics features and clinical data of simple repair patients before and after propensity score matching. The rate of male patients was similar between the groups as well as the BMI. The ASA score was significantly higher in the OpenA group [OpenA 123 patient (53.2%) vs LapA 95 patients (43.0%), p = 0.030]. No differences were found between the groups in terms of preoperative laboratory value (Hb, lactate, glycemia, WBC, PLT CRP) except for creatinine which was significantly higher in the OpenA patients (OpenA 1.3 ± 1.2 vs LapA 1.0 ± 0.7 p = 0.001). Regarding the scores that reflect patient general status such as shock index, CACI, EmSFI, and qSOFA only the last one was higher in the OpenA group than in the LapA group (qSofa: OpenA 64 ± 3.2 vs LapA 0.41 ± 2.60. p = 0.001). SIRS was anyway more frequent in the OpenA group [OpenA 108 patients (47.0%) vs LapA 71 patients (32.0%) p = 0.001]. Moreover, patients in the OpenA group were more fragile as reflected by a lower independency in daily activity and a higher EmSFI index. Operative details and postoperative outcomes before and after propensity score matching are summarized in Table 3.

Table 2 Demographics characteristics and clinical data of simple repair patients before and after propensity score matching
Full size table
Table 3 Operative details and postoperative outcomes before and after propensity score matching
Full size table

Boeys score, MPI and Jabalpur score were not different between-groups while Pulp score was slightly lower in the LapA group (OpenA 5 ± 3.4 vs LapA 4 ± 3.0). The mean operative time was 86.1 ± 31.3 min in the LapA group and 98.3 ± 40.1 in the OpenA group. The difference was statistically significant (p = 0.001). Mean diameter of the perforation and site were similar between the groups and no difference in terms of leak rate was retrieved. The overall morbidity rates was higher in the OpenA group [OpenA 93 patients (40.3%) vs LapA 49 patients (22.1%) p < 0.001] as well as the postoperative 30-day mortality rate [OpenA 45 patients (19.5%) vs LapA 19 patients (8.6%) (p = 0.001)]. As concern morbidity, MPI, Boey score, PULP score, and Jabalpur score showed no discrimination ability [AUC 0.491 (95% CI 0.433–0.548), 0.523 (95% CI 0.468–0.578), 0.562 (95% CI 0.507–0.617), and 0.501 (95% CI 0.442–0.560), respectively; p = 0.174] (Fig. 2). As regard to mortality Boey score e Jabalpur score exhibited poor discrimination ability [AUC 0.663 (95% CI 0.592–0.733), 0.604 (95% CI 0.528–0.680), respectively] while the accuracy of PULP score and MPI was acceptable [AUC 0.860 (95% CI 0.817–0.904), 0.740 (95% CI 0.674–0.807), respectively; p < 0.001] (Fig. 3). After propensity score matching, 172 patients were included in each group (the LapA and the OpenA). The analysis revealed that the pre-operative variables found to be significantly different before matching (i.e., age, ASA, CACI, EmSFI) were then well-balanced between both groups (Table 2) except for qSofa and altered autonomy, which have remained higher in the OpenA group. Outcomes of the propensity score matching demonstrated increased operative times in the OpenA (OpenA: 96.4 ± 37.2 vs LapA 88.47 ± 33 min, p = 0.035), with shorter overall length of stay in the LapA group (OpenA 13 ± 12 vs LapA 10.3 ± 11.4 days p = 0.038), and more frequent discharge to home (OpenA 81.3% vs LapA 72.6%, p = 0.017) (Table 3). Moreover, there was no statistically significant difference in mortality [OpenA 26 (15.1%) vs LapA 18 (10.5%), p = 0.258]. While focusing on morbidity, the overall rate of 30-day postoperative morbidities was significantly lower in the LapA group than in the OpenA group [OpenA 67 patients (39.0%) vs LapA 37 patients (21.5%) p = 0.002]. When stratified using the Clavien–Dindo classification, the severity of postoperative complications was statistically different only for C–D 1–2. However, it is important to note that when C–D 4 was taken into account the statistical analysis showed an Odds Ratio for OpenA almost three times [OR 2.870; (95% Conf. Interval) 0.888–9.265]. The results of regression analysis concerning suture leak and conversion to open surgery are shown in Table 4. Multivariate analysis revealed that pyloric/duodenal site, Boey score, and Age-Shock Index were the variables statistically related to suture leak while SIRS, platelet, serum lactate, MPI, and dimension were significantly associated with conversion to open surgery.

Fig. 2
figure 2

Comparison of receiver-operating characteristic (ROC) curve of MPI, Boey score, PULP score and Jabalpur score for overall morbidity

Full size image
Fig. 3
figure 3

Comparison of receiver-operating characteristic (ROC) curve of MPI, Boey score, PULP score and Jabalpur score for mortality

Full size image
Table 4 Univariate and multivariate analysis for Leakage and for Conversion to open surgery
Full size table

Discussion

The surgeons’ greater experience in laparoscopic surgery and the remarkable development of various laparoscopic surgical equipment and new anaesthesiology techniques had led to the improvement of the safety and feasibility of laparoscopic procedures. As a consequence, a large diffusion of minimally invasive approaches for the execution of more complex and demanding operations has been seen also in the emergency setting [18, 29, 30]. The surgical approach to perforated peptic ulcer has changed dramatically in recent decades. In the 1970s, vagotomy and pyloroplasty, with the whole risks associated with this type of surgery, was the procedure more frequently performed. In the following decade, a less aggressive approach was recommended: simple suture with or without omentoplasty [31]. In the last decades, many studies have reported the safety and the efficacy of LA for the treatment of PPU [16, 32]. Nevertheless, clinical reality differs from the results of randomised studies by a complex series of non-objectionable real-world data influencing treatment plans. In light of this, it is pivotal to define which patients are suitable for a laparoscopic approach. Laparoscopic repair of a PPU was described as early as 1990, but rates of adoption of this approach have been un-clear. Some international studies have reported a laparoscopic repair rate ranging from 41 to 76% and our series resulted in line with this findings [33, 34]. Moreover, in the 2020 WSES guidelines, the laparoscopic approach is suggested to be the first-line treatment for stable patients with small ulcers as long as surgeons are familiar with the skill and appropriate equipment is available [35]. In clinical practice, the patients who undergo a laparoscopic approach compared to an open approach have different characteristics, in terms of clinical presentation and comorbidities [36,37,38,39]. Our study shows that LapA patients are less fragile (lower CACI and EmSFI) and have a better overall general and inflammatory status (i.e., lower qSofa, lower rate of SIRS, and lower MPI). However, after propensity model only altered autonomy and qSofa were found to be statistically higher in the OpenA group. The main clinical scoring systems used to predict morbidity and mortality in PPU patients are Boey score, PULP Score and Jabalpur score [37, 38]. They are often used in conjunction with the Mannheim Peritonitis Index and ASA for improved comparison of the severity of physiologic derangement [38, 40, 41]. The Boey score seems to be the most disease-specific and is simple to calculate taking into consideration major medical illness, preoperative shock, and duration of perforation longer than 24 h before surgery. It is reported that a Boey score of 2 or higher indicates mortality rate greater than 30% [42,43,44]. Similarly to what performed by other authors, having many collected data, we retrospective calculated several common scores. The overall mean of the scores was similar between the OpenA and LapA groups both before and after propensity score matching. A large English population-based cohort study of Leusink et al. confirmed that laparoscopic repair of PPU is associated with a significant reduction in 30-day and 90-day mortality, postoperative pneumonia, and length of hospital stay [15]. A recent meta-analysis of randomized clinical trials comparing laparotomy to laparoscopic repair found no difference in morbidity or mortality between the two procedures [45]. Although previous cohort studies have compared outcomes between LA and OA, the matched nature of this study makes the findings more informative for the surgeon's daily activities. We found that the benefits of a laparoscopic approach include a shorter operative time and length of stay and a slightly lower mortality rate. According to the literature, perforation carries an associated mortality rate of 10–40%; our overall mortality rate is approaching the lower limit and we evidenced that the rate of death after laparoscopic surgery was similar to the rate of the open surgery. These findings are in line with the literature: a recent Danish propensity analysis with a large study population failed to demonstrate a lower mortality in the laparoscopic group [46]. Only one English retrospective population-based study, and a meta-analysis with non-randomized studies included in the analysis, were able to demonstrate the benefits of laparoscopy in terms of mortality [15]. In a meta-analysis, Zhou et al. found significant differences in hospital mortality between the laparoscopic repair and open repair groups in the high quality non-randomized studies, but not in the clinical trials [47]. Like Mirabella et al., we believe that mortality depends more on the risk factors of the patient and the aggressiveness of the ulcer than on the surgical approach [48]. Focusing on morbidity, literature reports that postoperative complications usually occur in 30% of cases. Our morbidity rate is in line with the literature with an overall morbidity rate of 30.2% that drops to 21% in case of laparoscopic approach with simple repair. Moreover, the severity of postoperative complications was statistically different only for C–D 1–2, but it is important to note that when C–D 4 was taken into account the Odds Ratio for OpenA group was almost three times. The predictive value of morbidity of all the analysed scores was very low, equivalent to chance. On the contrary, when the accuracy of the scores was evaluated in predicting mortality, our series showed that PULP score and MPI had acceptable accuracy with the former having the highest AUC value. The main cause for reoperation following surgical repair is suture leak. Proposed explanations, based on current literature data, include the difficulty in laparoscopic knot tying, ulcer diameter (> 2 cm) and abdominal contamination [49, 50]. Based on our multivariate analysis the factors associated with leak were ulcer site (pyloric/duodenal), higher Boey Score and higher Age-shock index while PULP score revealed his statistically significant association with leakage only at the univariate analysis. A number of studies have reported the utility of Boey score and PULP score in predicting conversion of laparoscopic PPU repair while Muller et al. stated that the conversion to an open approach could only be assessed intraoperatively [51, 52]. In our series, as regards conversion to open surgery, PULP score and Jabalpur score were identified as risk factors only at univariate analysis while Boey score was not associated with conversion. Furthermore, our multivariate analysis showed that SIRS, platelet, serum lactate, MPI, and ulcer dimension were statistically risk factors and this is unanimous with the literature where is reported that the main reasons for conversion are a posterior location that did not allow a proper inspection of the ulcer, a large perforation size (> 2 cm), severe inflammatory involvement of the surrounding tissue, adhesions, and suspected tumours [51, 53]. In light of this, investigating our series, we were not able to determine the actual causes which required conversion to laparotomy because the data collected did not include enough information about this item. However, a brief survey allowed us to detect that the main cause of conversion was the ulcer size followed by anaesthesiologist decision for medical reasons and by technical difficulty due to adhesions. Consequently, we agree that a conversion to an open procedure should only be assessed intraoperatively.

Limitations

As it clearly derives from above, this study has several limitations. First, it is a mixed prospective and retrospective multi-centre study with not previously established common treatment protocol. Although data were collected following a homogeneous method, decisions regarding the timing and the choice of approach might differ among institutions or even in the same hospital due to attending surgeon preference and expertise or intra- and inter-hospital settings and instrumental availability in not a referral centre or after-hours. In addition, the suture technique could be slightly different by the surgeon’s preference. Therefore, the analysis was not stratified by participating surgeons or institutions because the protocol did not enable a separate comparison of data for obvious professional ethical reasons. Another issue to take into account is that a part of the study included patients treated during the COVID-19 pandemic, which has dramatically altered the clinical practice also in emergency surgery. This is why in this study there is a high risk of patient selection bias. However, the number of cases recruited along with the large amount of collected data provided a good sample size for multiple logistic analyses mitigating these limitations. Moreover, a multicentre study allows better generalisation of results than a single centre, while the propensity score model let us to compare two similar restricted groups excluding confounders and correcting for undetectable selection bias. Furthermore, in order to evaluate the potential benefits of the laparoscopic approach, it is remarkable that postoperative pain was not assessed because medical records did not always provide this data. Lastly, no details about long-term outcomes such as incisional hernia or recurrence of PPU or bleeding were evaluated.

Conclusion

Despite the promising results of laparoscopy and the suggestion of literature guidelines, the open approach is still widely adopted as our research also witnesses. Based on the results of the present study, we can support that laparoscopic suturing of perforated peptic ulcers, apart from being a safe technique, could provide significant advantages in terms of postoperative complications and hospital stay showing its non-inferiority when compared with open approach. Therefore, we advocate that laparoscopy should be always used as first choice in stable patients in order to confirm the diagnosis, localise the site and size of the perforation, determine the extent of peritonitis, and try to perform suture even with the use of barbed knotless suture.