Introduction
Colorectal cancer is the third most common cancer worldwide, with the majority of patients treated with surgical resection and primary anastomosis. Between 1% and 20% of colorectal resections result in an anastomotic leak (AL) [
1], which poses a significant risk for patients. As well as the immediate threat of developing faecal peritonitis and sepsis, with its attendant morbidity and mortality profile, in the long term AL is associated with a lower 5-year cancer-specific survival rates [
2‐
7] and increased risk of cancer recurrence [
2‐
4,
8,
9]. Hence, there is an urgent and unmet need to identify AL quickly to enable early intervention, possible anastomotic salvage and avoid long-term complications.
Current methods of diagnosing AL are usually based on serial clinical examinations and radiological imaging from days 3–5 postoperatively. These include symptoms such as pain, tachycardia, fever, oliguria, ileus, diarrhoea and leukocytosis, in combination with elevated blood levels of C-reactive protein (CRP). However, the clinical signs are non-specific and are only observed once there is a systemic response to the AL. Similarly, blood CRP can aid diagnosis, with postoperative day 3 levels of less than 172 mg/l having a negative predictive value of 97% but only a low specificity for positive diagnosis of an AL [
10].
By postoperative day 3 to 5 the risk of mortality increases, alongside the necessity for an emergency operation to treat faecal peritonitis with its concomitant high risk of a permanent stoma formation. Earlier detection of an AL could change the management and treatment of these patients and, crucially, improve their postoperative outcome and quality of life.
In the present study, we examined whether peritoneal fluid drained from the surgical site can be used to diagnose AL more rapidly and accurately than current practice. By assaying peritoneal fluid in proximity to the anastomotic site, one should be able to detect the local response to colonic contamination of the peritoneum more quickly, sensitively and specifically than later systemic responses. This hypothesis was based on previous studies in which peritoneal fluid was found to respond more rapidly to the presence of AL than routinely sampled blood [
11], confirming that it is more likely to be an effective medium for robust early detection. Similarly, our own research in different pathologies has shown that local immune signatures (“immune fingerprints”) at inflammatory sites are powerful predictors of acute infections, including in the peritoneal cavity of patients receiving peritoneal dialysis [
12,
13], in the urine of patients presenting with suspected urinary tract infection [
14], and in cerebrospinal fluid of neurosurgical patients [
15]. Such disease-specific immune fingerprints combining clinical parameters with soluble and cellular biomarkers of inflammation, organ damage and/or physiology are likely to improve on the reliance on a single biomarker and increase both the sensitivity and specificity of early diagnostic tests.
In order to have immediate relevance for patient benefit, the present study was designed to assess the practical utility of biomarkers in drain fluid after abdominal surgery and test a suite of clinically approved biomarkers using the Siemens IMMULITE platform, an automated immunoassay system routinely used for quantifying cytokines in hospitals globally.
Methods
Ethics approval
All methods were carried out in accordance with relevant guidelines and regulations. Experimental protocols were approved by Cardiff University, and written informed consent was obtained from all subjects. Recruitment of patients was approved by the Wales Cancer Bank under reference no. 18/016 and conducted according to the principles expressed in the Declaration of Helsinki.
Objectives
The primary objective of the study was to assess whether altered levels of peritoneal fluid biomarkers (cytokines) in response to AL could be measured on a widely available commercial immunoanalyser platform.
The secondary outcome was the analysis of peritoneal biomarker measurements to explore their sensitivity and specificity for AL diagnosis.
Patient cohort
Between June 2019 and June 2021, all patients undergoing an elective anterior resection within the colorectal department of the University Hospital of Wales in Cardiff (UK) were screened for inclusion. Inclusion criteria were cancer diagnosis, intended primary anastomosis, age ≥ 18 years and capacity to consent. Exclusion criteria were non-cancer diagnosis, lack of capacity and a permanent end colostomy formation. Patients who had neoadjuvant chemoradiotherapy were included, and the intention to form a defunctioning ileostomy was not an exclusion criterion. Operations were performed according to surgeon preference, either laparoscopically or via a midline laparotomy. Laparoscopic procedures had extraction sites via a midline or a Pfannenstiel incision. The primary anastomosis was formed using a conventional circular stapler, the size according to surgeon choice. At the end of the operation a non-suction drain was placed in the left iliac fossa, into the pelvis. The removal of the drain was decided upon by the operating surgeon but was left in situ for at least 48 h postoperatively.
Clinical assessments
Patient demographics were collected preoperatively. These included age, sex, body mass index, smoking status, and previous chemotherapy and/or radiotherapy. Intraoperative variables were also collected, including operation time, laparoscopic or open approach, and formation of defunctioning stoma. Postoperatively, data were collected on length of hospital stay, tumour staging (TNM) and postoperative complications. Complications were classified as AL or other complications, which included bleeding, small bowel obstruction and any other reason for return to theatre. AL was defined as clinically manifest insufficiency of the anastomosis leading to a clinical state requiring treatment (i.e. grade B/C) [
16]. AL was confirmed by either computer tomography (CT) scan and/or reoperation. As such, patients were stratified into three groups for analysis: ‘uneventful recovery’, ‘anastomotic leak’ and ‘other complications’. All clinical data were incorporated into a database in which data were pseudo-anonymised by assigning each patient a study number. No personally identifiable information was included.
Collection and storage of peritoneal fluid
Drain peritoneal fluid was collected at 8:00 am on days 1 and 2 postoperatively, as well as on day 3 if that patient still had the peritoneal drain in situ. The first sample (day 1) was collected between 14 and 20 h postoperatively. All samples were transferred for analysis in the laboratory within 30 min. The peritoneal fluid was centrifuged twice at 500×g for 20 min at 20 °C. The cell-free supernatant was removed, aliquoted into 2-ml cryotubes and stored at − 80 °C until further use.
Peritoneal drain fluid analysis
Peritoneal fluid levels of interleukin (IL)-1β, IL-6, IL-8/CXCL8, IL-10, tumour necrosis factor alpha (TNFα) and CRP were measured using the corresponding IMMULITE 1000 kits on IMMULITE 1000 Immunoassay systems through the solid-phase chemiluminescent immunometric assay (CLIA) method. All kits and instruments were supplied by Siemens Healthineers GmbH, Germany. Normal assay ranges were as follows: IL-1β, 5–1000 pg/ml; IL-6, 2–1000 pg/ml; CXCL8, 5–7500 pg/ml; IL-10, 5–1000 pg/ml; TNFα, 4–1000 pg/ml; and CRP, 0.3–100 mg/l. For CRP measurements, samples were prediluted 1 in 100 in a CRP-free protein/buffer matrix prior to assay. As a result of inherent high concentrations of IL-6 in peritoneal fluid, two dilutions (1 in 50 and 1 in 100) were used to cover the reasonable expectations of dilutional requirements, using an IL-6-free protein/buffer matrix prior to assay. All other biomarkers were measured directly in the undiluted fluid.
To evaluate precision, single donor or pooled peritoneal fluid samples were selected and prepared at different concentrations, divided into aliquots, and stored at − 20 °C until use, with a fresh aliquot used for each run. For IL-1β, IL-6 and IL-10, repeatability/within-lab precision was measured by running the prepared pools once per day, for 5 days, in replicates of five. For CRP, TNFα and CXCL8, repeatability was assessed from 10 replicates run from the same sample aliquot in a single run. Linearity was assessed by diluting a peritoneal fluid pool with an analyte concentration near the upper reportable limit (or as high as the sample pool would allow), with a peritoneal fluid pool containing low concentrations of the relevant analyte to yield concentrations of 100%, 75%, 50%, 25%, 10%, 5%, 2.5%, 1%, and 0% of the original high-concentration pool. Separate pools were prepared for each analyte. All dilutions were tested in duplicate. Linearity of IMMULITE assays for peritoneal drain fluid was then analysed by weighted regression. To evaluate recovery of cytokines, peritoneal fluid samples were spiked with 2–3 solutions containing different amounts of each respective analyte. Samples for determination of recovery were prepared by making stock solutions in analyte-free protein buffer matrix. Stock serum solutions were added to peritoneal fluid samples at a ratio of 1:20. Spiked peritoneal fluid samples were assayed in duplicate, and the ratios between observed and expected concentrations were calculated.
Data analysis
Data were analysed in R4.1 (R core team, 2017) [
17] and graphed in Prism 9 (GraphPad Software, San Diego, California, USA). Biomarker thresholds were determined by maximising F1 scores across all timepoints. The F1 score is the harmonic mean between precision and recall, and is optimal for datasets with a large proportion of true negative patients. To create the 6C metric, log2 concentrations in pg/ml (CRP) or ng/ml (IL-6) were combined into an unweighted linear metric, with the boundary optimised by F1 score. Biomarker weighting was tested but found to not improve accuracy.
Discussion
For patients undergoing a surgical colorectal resection, AL is a significant complication that changes the postoperative pathway, with increased short-term morbidity and mortality, and long-term reduced cancer-specific survival. Analysis of the peritoneal fluid of the patients included in this study has demonstrated that inflammatory biomarkers are both measurable on a commercial platform and could be accurate in diagnosing an AL in the early postoperative period. Whilst this study has utilised the IMMULITE platform for the drain fluid testing, the two key markers of interest are available on other automated immunoassay platforms, such as Roche Elecsys, Beckman Coulter Access, widely available and routinely used in hospital clinical laboratories. This is the first study to show that drain fluid could be utilised in these widely available immunoassays with minimal impact on overall analytical performance, hence demonstrating the analysis and interpretation of peritoneal fluid could be comparable to blood tests. Despite the samples being frozen and interpreted as a group in this study, in clinical practice the samples would not need to be frozen and could be analysed immediately as required by hospital laboratories, further adding to the clinical applicability of this study.
The use of a peritoneal drain is still a variable practice amongst colorectal surgeons and departments particularly since the widespread adoption of ERAS principles [
18]. Previous studies, including a multicentre European collaborative study, have demonstrated that the placement of a peritoneal drain did not reduce the rate of AL or allow earlier detection. The use of drains was not associated with major complications, but did increase the rates of surgical site infections and increased the length of stay [
19,
20]. Each study has focused on drain placement as a method of preventing ALs; however, this pilot study demonstrates that a short-term placement of a peritoneal drain (48 h) could provide accurate diagnostic capabilities, thus justifying their placement. There is further data that a peritoneal microcatheter could be used in place of a drain, which could potentially minimise the adverse effects including wound infections and impact on ERAS implementation [
21].
All biomarkers tested, with the exception of CXCL8, showed a measurable increased level in peritoneal fluid in those patients with AL at day 3, in agreement with previous published studies [
11,
22]. IL-6 is an inflammatory cytokine which is upregulated in response to microbial pathogens but also tissue injury. Consequently, when measured in blood it is a non-specific marker of systemic inflammation [
23]. Peritoneal IL-6 has been measured before in AL studies and has been proposed as a predictive marker for the presence of AL in patients after surgery in isolation [
24]. It is broadly successful, with sensitivities and specificities in line with our measurements, but is insufficiently accurate to be used in isolation. In contrast, the present study shows that it can be combined specifically with local CRP levels derived from drain fluid to create an effective metric that can detect AL within 48 h postoperatively with high positive and negative predictive values.
CRP is commonly measured in blood where it is non-specifically increased in inflammatory and infective conditions, including in postoperative patients with no complications. CRP can act as an antibacterial effector which binds surface polysaccharides, opsonising bacteria including those that would be released into the peritoneum when an AL occurs, and activating the classical complement cascade. Systemic levels of CRP can be used as a negative predictive value of AL between 3 and 5 days postoperatively [
10,
25], in keeping with its responsiveness to inflammation. CRP is commonly characterised as an acute phase protein made by hepatic cells in response to IL-6 and IL-1β [
26]; however, production can also be stimulated in macrophages [
27]. Our detection of elevated CRP in drain fluid 48 h postoperatively supports the hypothesis that it is produced locally in response to leakage of colonic content. Importantly, the differences in drain fluid CRP levels between patients with AL and those experiencing inflammation due to other complications at this time point show that measuring local CRP has the potential to be a more specific biomarker for AL than systemic CRP levels.
Dual detection of blood levels of IL-6 and CRP has already been proposed as a combined biomarker of AL [
28]. However, while both markers have excellent negative predictive values, leading to high area under the receiver operating characteristic curve (AUROC) scores, positive predictive values for AL have remained poor so far. Our current study shows that sampling of drain fluid has both greater sensitivity for AL detection than blood markers and gives reliable results 24 h earlier. The magnitude and early detection of the inflammatory response likely reflect the proximity of the fluid to the site of the inflammatory event, which is in agreement with studies that have shown IL-6 release from peritoneal membranes within 2 h of surgery [
29]. Interestingly, IL-10, which we found at levels averaging 200 times lower than IL-6, was below the level of detection in the same study. The use of drain fluid rather than blood could also potentially mitigate the general nature of IL-6 and CRP being increased in response to unrelated inflammatory stimuli, such as concurrent infections and systemic conditions. This has been shown to be the case in other contexts such as meningitis, where it was found that blood IL-6 was not as specific a marker as IL-6 in the cerebrospinal fluid [
15].
While promising, this is a pilot study and needs to be replicated with more samples in a larger study. Given the distribution of the readings, it is possible that in a larger cohort IL-10 or IL-1β may also prove to be useful in a combined metric with IL-6, and investigating whether these could add to the robustness of the metric will be an important next step to the research. The patient cohort measured in the present study was also limited to those undergoing an anterior resection, and results were not stratified on tumour location in relation to the peritoneal fold or the distance of tumour/anastomosis from the anal verge. Without the inclusion of all pathologies (e.g. cancer and inflammatory bowel disease) and all colonic resections, it is difficult to assess the applicability of the results to all types of anastomosis and patients. However, this is the first study to indicate a sensitive and specific marker for AL before clinical manifestation.
The impact of an earlier diagnosis of an AL is hard to predict at present, and further research is needed into the best management strategy for patients who are predicted to develop an AL. However, an accurate method to diagnose and thus the ability to intervene and reduce the need for a defunctioning ileostomy, with its associated morbidity and need for reversal, is likely to ameliorate the risk of AL-related complications before the development of systemic sepsis and improve the patient’s pathway.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.