Introduction

Very few data are available concerning the clinical characteristics of postoperative infections following bariatric surgery (BS) in patients admitted to intensive care unit (ICU) [1]. From the ICU physician's standpoint, this subset of patients represents new and challenging issues and raises specific concerns. In a group of patients referred to our ICU with a diagnosis of postoperative peritonitis (POP), we have previously reported delayed identification of complications, frequently at the time of organ dysfunction [2]. In order to identify the most relevant features in these patients, we conducted a prospective observational study to compare the clinical characteristics and outcome of a group of patients admitted to our ICU for POP following BS (called BS cases) and a group of patients who developed POP following conventional surgery (called cPOP).

Patients and Methods

Patient Selection

The study was approved by the local institutional review board (CEERB CHU Bichat Paris VII University, Paris), which waived the need for signed informed consent. All patients hospitalized in our ICU for POP between 2001 and 2011 were included in a prospective cohort database. The diagnosis of POP was systematically assessed by reoperation and was based on intraoperative findings, as previously described [3]. Exclusion criteria were absence of intraoperative samples or negative peritoneal culture. Source control was achieved in every case.

Morbidly obese patients (body mass index (BMI) >40 kg/m2 or >35 kg/m2 with underlying disease related to overweight) underwent restrictive procedures (adjustable gastric banding (LABG) or sleeve gastrectomy (LSG)) or Roux-en-Y gastric bypass (RYGB).

Susceptibility Testing and Empiric Antimicrobial Therapy

Peritoneal fluid samples were systematically collected during reoperation [3]. Gram stain for direct examination and cultures were performed with identification and susceptibility testing. Blood cultures were drawn during the 24 h preceding and following reoperation. Antibiotic susceptibility was determined by the disk diffusion method [4]. Multidrug-resistant bacteria (MDR) were defined as previously described [3]. Broad-spectrum antibiotics were arbitrarily defined as piperacillin/tazobactam (pip/taz), imipenem (imi), and fluoroquinolones [3]. Interval antibiotics were defined as agents administered between the initial surgery and reoperation, for at least 24 h and started at least 24 h before reoperation.

Our institutional protocol for parenteral empiric antibiotic therapy (EAT) complies with the national [5] and international [6] guidelines for the management of intraabdominal infections. EAT is systematically initiated at the time of reoperation, takes into account the severity of the case, and usually combines pip/taz or imi with amikacin and vancomycin possibly combined with antifungal therapy based on presumed risk factors [7].

Data Collection

Demographic data, underlying medical condition [8], characteristics of the initial surgery, antibiotic regimen, and interval antibiotics were collected on ICU admission. In patients who underwent BS, BMI was recorded. The Acute Physiology and Chronic Health Evaluation II (APACHE II) [9] and the Sequential Organ Failure Assessment (SOFA) [10] were calculated on the day of reoperation for POP. Organ failure (grade ≥3 of the SOFA score) of each organ system was reported. The etiology of POP was noted. Identification and susceptibility testing of pathogens in blood cultures and peritoneal fluid were recorded.

The type and duration of EAT and definitive antibiotic therapy based on microbiologic results were recorded. Adequacy of EAT was assessed according to susceptibility testing, corresponding to all bacteria isolated susceptible to at least one antibiotic administered.

De-escalation was defined as either discontinuation of an antimicrobial agent or a change from one antibiotic to another, while escalation was defined as the addition of a new anti-infectious agent or a change of antibiotic therapy in the opposite direction.

ICU Outcome

Time to and number of reoperations, duration of mechanical ventilation, length of ICU stay, and ICU and in-hospital mortality were recorded.

Statistical Analysis

Results are expressed as mean ± one standard deviation (SD) or number and proportions. Statistical analysis used R software (R Foundation for Statistical Computing, Vienna, Austria). The BS group was compared with the cPOP group using Student t-test and Wilcoxon unpaired test for continuous variables and chi-square test and Fisher's exact test for discrete variables. A secondary exploratory within-group analysis was also performed. Firstly, in the cPOP group, characteristics were compared between operations performed above and below the transverse mesocolon. Secondly, in the BS group, characteristics were compared between restrictive surgery and RYGB procedures. Due to its exploratory nature, this secondary analysis did not intend to draw any definitive conclusions.

Risk factors for mortality were analyzed. ICU survivors and non-survivors were compared by univariate analysis. Odds ratio (ORs) and 95 % confidence intervals (CI) were calculated. Variables with a p-value <0.2 on univariate analysis were entered into a multivariate logistic regression analysis with backward selection. Variables with a p-value ≥0.5 at each step of the regression analysis were rejected. Interactions between final variables were tested and significant interactions (p < 0.05) were entered in the model. The final model with interactions was evaluated for performance with Nagelkerke's R 2, for discrimination with the c-statistic, and for calibration with Hosmer–Lemeshow's test. Statistical significance was defined as p < 0.05.

Results

Study Population

Overall, 183 patients with POP were admitted to our ICU, of which 32 patients (65 %) in the BS group and 49 patients (29 %) in the cPOP group underwent initial surgery in another institution (p < 0.0001). During the study period, 554 adjustable gastric bandings, 603 sleeve gastrectomies, and 1,608 RYGB were performed at our institution, with three deaths after sleeve gastrectomy and RYGB.

Clinical characteristics on admission and at the time of POP are shown in Table 1. The initial BS consisted of 15 LAGB, eight LSG, and 26 RYGB. Clinical presentations were similar in BS patients who underwent restrictive surgery or RYGB, except for a trend toward increased proportions of renal failure in patients who underwent RYGB (p = 0.055) (Table 1). The main surgical findings at reoperation in the BS patients are displayed in Table 1. In the group of restrictive procedures, six anastomotic leaks were identified among patients with LSG (four repaired by suture and lavage/drainage and one treated by drainage and endoscopic stent; one patient required total gastrectomy with secondary anastomosis and feeding jejunostomy), and 13 gastric perforations were detected in patients after LAGB, which required removal of the band with suture of the perforation and lavage and drainage. In the RYGB group, the majority of leaks involved the gastrojejunal anastomosis. Overall, resection of the anastomosis with primary reconstruction or delayed esophagojejunostomy was performed. One patient presented stenosis of the jejuno-jejunal anastomosis, causing perforation of the gastric remnant, which was treated by revision of the anastomosis and total gastrectomy.

Table 1 Main characteristics of the two groups of patients with postoperative peritonitis following conventional and bariatric surgery expressed as mean ± standard deviation (SD) or number (n) and proportions
Full size table

Microbiologic Results

Bacteremia was reported in eight (16 %) BS patients (including four streptococci and three Enterobacteriaceae) and 33 (25 %) cPOP patients (including eight anaerobes, seven Enterobacter spp, and six Escherichia coli (n = 6)).

A total of 122 micro-organisms were cultured from the peritoneal fluid of BS patients, and 402 micro-organisms were cultured from cPOP patients with a mean number of 2.5 ± 1.2 versus 2.9 ± 1.4 micro-organisms, respectively (p = 0.03) (Table 2). Gram-positive strains represented 63 % of monomicrobial isolates of the BS group. Mixed infections involving Gram-negative/Gram-positive samples were observed in 12 (24 %) patients from the BS group compared to 64 (48 %) cPOP cases (p = 0.006) (Table 2).

Table 2 Micro-organisms isolated from peritoneal fluid (expressed as total number and proportions) in the two groups of patients with postoperative peritonitis following conventional and bariatric surgery
Full size table

Eleven MDR strains were cultured from the 11 BS patients (eight Gram-positive including seven methicillin-resistant coagulase negative staphylococci (MRCNS) and two Enterobacteriaceae). A total of 73 MDR strains (35 Gram-positive and 38 Gram-negative organisms) were cultured in the cPOP group (one strain in 43 patients and ≥2 strains in 15) including MRCNS (n = 24), methicillin-resistant Staphylococcus aureus (n = 6), Enterobacter spp (n = 13), Pseudomonas aeruginosa (n = 9), and E. coli (n = 7).

Candida albicans (n = 11) was the predominant fungus cultured in the BS group. Susceptibility to fluconazole was confirmed in 100 % (6/6) of evaluated strains. In the cPOP group, the most frequent fungi were C. albicans cultured in 23 (50 % of fungi) cases (12 strains susceptible to fluconazole) and Candida glabrata in 13 (28 %) cases (3/6 strains resistant to fluconazole).

Antibiotic Therapy

At the time of reoperation for POP, EAT was initiated in all cases (Table 3). Monotherapy was administered to only a limited number of cases and mainly consisted of pip/taz (six BS patients and 32 cPOP patients). A total of 84 organisms were not adequately treated by EAT. Inadequacy of combination therapy was observed in seven (14 %) BS patients and 32 (24 %) cPOP patients. Presence of MDR strains was an important factor of inadequacy: six MDR strains were reported in six of the nine (66 %) BS patients and 41 MDR strains were reported in 32 (66 %) cPOP patients with inadequate EA therapy.

Table 3 Main agents used as empiric antibiotic therapy and definitive treatment in the two groups of patients with postoperative peritonitis following conventional and bariatric surgery expressed as number and proportions
Full size table

Poor clinical response and/or susceptibility testing led to escalation therapy in three (6 %) BS patients versus 31 (23 %) cPOP patients (p = 0.009) at the time of definitive antibiotic therapy. De-escalation was performed in 33 (67 %) BS cases and 69 (51 %) cPOP cases (p = 0.06).

At the time of definitive therapy, monotherapy mainly consisted of amoxicillin/clavulanic acid in BS patients (12 (24 %) patients), while in the cPOP group, monotherapy was based on amox/clav (n = 15, 11 %), pip/taz (n = 14, 10 %), or imi (n = 8, 6 %). Overall, amox/clav represented 34 (69 %) of all definitive therapies in BS cases. Antifungal therapy, mainly fluconazole, was given in 21 (43 %) BS patients and 46 (34 %) cPOP cases (Table 3).

Patient Outcome

Following the first reoperation in our institution, 19 (39 %) BS patients and 75 (56 %) cPOP patients had a favorable outcome (p = 0.04) (Table 4). A total of 12 (24 %) BS patients and 43 (32 %) cPOP patients died (p = 0.32) after a mean interval of 17 ± 16 vs 21 ± 16 days, respectively (p = 0.37). No significant difference was observed between BS patients treated by restrictive surgery or RYGB.

Table 4 Mortality and morbidity criteria in patients with postoperative peritonitis following conventional and bariatric surgery. Results are expressed as mean ± standard deviation (SD) or number (n) and proportions
Full size table

The risk factors for death identified on univariate analysis are presented in Table 5. The BS variable was forced into the multivariate logistic regression analysis model. No significant interaction was observed between the final variables. On multivariate analysis, four criteria were significantly associated with outcome, but none of them were specifically linked to BS (Table 6).

Table 5 Univariate analysis of outcome at discharge from intensive care unit, expressed as mean ± standard deviation or total number (%)
Full size table
Table 6 Multivariate analysis (with backward selection but similar results with forward and stepwise selection)
Full size table

Discussion

To the best of our knowledge, our results provide a unique picture of patients who develop POP following BS. These BS patients represent an atypical group that differs from cases of cPOP or cannot be considered to be similar to patients with complicated upper mesocolic surgery. Our highly selected population and the observational study design could be considered to be poorly representative of the complications of BS as most of these complicated cases are treated in surgical wards. We nevertheless assume that this global analysis could provide general conclusions that could be applied to all complicated cases of BS in the ICU setting, although RYGB appears to present a number of specificities. In addition, our recruitment does not reflect the local incidence of these complications. Consequently, we cannot provide a clear estimate of the proportion of BS patients requiring ICU admission.

In studies reporting cPOP patients admitted to the ICU, the proportions of upper mesocolic cases usually ranged between 13 and 30 % of cases [11–14]. While leak rates are close to zero in the postoperative course of gastric banding [15–17], rates as high as 20 % have been observed after RYGB. Recent publications report mortality rates after anastomotic leaks ranging between 6 and 22 % [18–20] and up to 66 % in some series [21, 22] of RYGB. The delayed diagnosis of these leaks following BS ranges between 1 and 18 days [23] as they are frequently asymptomatic or associated with limited clinical signs [18, 23, 24]. This paucity of symptoms has been previously reported in ICU cases [2]. The large number of reoperations performed before transfer to our institution suggests that the postoperative complication was diagnosed early in many cases. However, technical issues and recurrent or persisting infections justified transfer of a large number of cases.

Polymicrobial infections and prior antibiotic usage are common in the course of POP and may create a bias in the samples retrieved. The high frequency of Gram-positive cocci and fungi reported in BS cases is not surprising as it has been previously observed in patients with gastroduodenal or upper mesocolic perforations, including cPOP cases [25]. The microbiologic characteristics of patients who underwent complicated RYGB have not been previously studied. Our results suggest that these patients should be considered as having small bowel perforation rather than upper gastrointestinal perforation, but the overall ICU clinical management does not appear to be different from that applied to restrictive BS cases. High concentrations of aerobes, anaerobes, and fungi have been reported in the stomach flora of patients undergoing bypass for morbid obesity [26]. The high pH reported in these patients is also a predisposing factor to yeast growth [26]. The prior use of antibiotic therapy before the diagnosis of POP is a major risk factor for the emergence of MDR strains and fungi. This risk of MDR strains justifies the use of EAT protocols comprising broad-spectrum agents [3]. Interestingly, the overall proportion of MDR strains remains low in BS cases and allows de-escalation therapy in most cases [3]. These observations are helpful to guide EAT and suggest that complicated BS patients are at moderate risk for harboring MDR strains and at high risk of fungal infection. Based on the candida peritonitis score [7], our local protocol recommends early antifungal therapy in these patients. This combined approach allows an acceptable adequacy of 80 % of all empiric therapy in BS cases, much higher than that observed in cPOP patients.

Recent studies have reported contradictory results on the prognosis of morbidly obese patients admitted to the ICU, from a decreased risk of in-hospital mortality in surgical patients [27] to increased morbidity in a mixed population [28] and even an increased risk of death among surgical patients [29]. When compared to BS cases treated conservatively, in whom no death is reported, the patients who require reoperation have high mortality rates of up to 18 to 22 % [24, 30], and even to 40 % in some specific subpopulations [31]. Despite their younger age and a low incidence of underlying disease, our BS patients demonstrated similar levels of severity, and even higher initial rates of renal failure than cPOP patients. This point suggests that these patients should be considered to be at high risk of complications and death, although BS did not appear to be a significant risk factor for mortality on multivariate analysis.

Conclusion

The present data suggest that, despite a younger age, the severity of POP following BS results in high mortality rates, similar to those observed in older patients with cPOP. The usual empiric antibiotic therapy protocols should be applied because of the potential risk of MDR strains, but de-escalation can be performed in most cases.