International consensus statement regarding the use of animal models for research on anastomoses in the lower gastrointestinal tract

In December 2014, an extensive literature search was conducted using the PubMed (MEDLINE) OvidSP (EMBASE) databases for all papers related to animal models, which were used in an experimental setting to either investigate anastomotic healing or anastomotic leakage in the lower gastrointestinal tract. Search terms included: “anastomosis/anastomotic,” “leak/leakage,” “bowel/intestine/colon/colorectal,” and “animal/animals/rat/mouse/mice/pig/dog/goat/rabbit/animal model.” English and Dutch were used as language restrictions and the search was limited to articles published between January 1, 2000 and December 1, 2014. The following inclusion criteria were applied to the titles and abstracts of the search results: experimental setting, use of an animal model, and an anastomosis made in the lower gastrointestinal tract (GIT) (gastroduodenal/gastrojejunal were considered upper GIT and therefore excluded). We excluded commentary reports, review articles, and articles containing results that had been previously reported in another included article. All articles were combined in a single list of which JWAMB and LvB identified eligible reports; in case of discrepancy, agreement on inclusion was reached through discussion with MA-T as a third reviewer. For an overview of the study selection, performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline, see supplementary data figure S1.

Our primary outcomes were (1) type of animal used and (2) location and type of anastomosis. As secondary outcomes, we evaluated scoring models used for macroscopic findings reporting on AL, i.e., adhesions, bursting pressure, histology, and other examinations performed. Further, we assessed the country of origin where the study was performed and the year in which the study was published and how animal welfare was reported.

The qualitative review of the literature served as our starting point for the Delphi technique. The main goal was to achieve consensus on the use of animal models for research on anastomoses in the lower gastrointestinal tract, specifically on which animal, location, and type of surgery; macroscopic outcome; histological assessment; mechanical and biochemical outcome measures; and animal testing and welfare. The Delphi technique is a widely used and accepted consensus method for gathering data from respondents within their domain of expertise in order to formulate recommendations or guidelines that can be used in the future. For this report, we contacted all principal investigators from included articles by email, explained our study, and invited them to participate in this international Delphi project.

Questionnaires were developed and distributed using SurveyMonkey (SurveyMonkey Inc, Palo Alto, CA; www.​survey-monkey.​com). This online survey contained several questions on three main subtopics of intestinal anastomotic research: the first part consisted of questions regarding animal model used and reasoning for the choice of this model; the second part focused on macroscopically scoring and measurements performed on the anastomosis (leakage rate, adhesion evaluation), and the last part inquired about histological analysis and additional tests (e.g., bursting pressure, ELISA, quantitative polymerase chain reaction (qPCR)) that were used to gain more insight in the healing process (inflammation, proliferation). There were no open questions in the survey, but participants were encouraged to provide arguments for their choice, suggestions, or additional remarks in free text fields below each question.

After receiving participants’ responses, the collected information was converted into a second questionnaire. This round included the items and ratings summarized from the previous round. Here, we asked all participants to revise their judgments or specify reasons why they were not convinced of the most commonly used animal model. All items achieving consensus, remaining items, and their ratings, as well as minority opinions, were reported during the questionnaires. The RAND/UCLA Appropriate Method (RAM) [34] was used to assess consensus in an expert panel on the use of animal models, macroscopic scoring of leakage and adhesion, mechanical and biochemical parameters, and histological outcome. For reporting our study, we used the recommendations of Sinha et al. regarding the Delphi technique [35].

Results of the survey were exported to MS Excel 2011 (Microsoft Corp, Redmond WA). Consensus was reached if the panel rated the subject inappropriate (panel median 1–3) or appropriate (panel median 7–9) on a 9-point scale without disagreement, according to the method used by Moossdorff et al. [36]. Disagreement was tested using the interpercentile range adjusted for symmetry in accordance to the RAND/UCLA Appropriateness Method Manual [34].

In total, 277 articles were retrieved from the search. After screening, 167 articles were included in the systematic review (for flowchart see supplementary data S1). An ongoing increase in publication frequency on anastomotic healing and leakage was found, from only 3 in 2000 to 18 in 2014 (Fig. 1a). Animal models used in these experiments were rat (65 %), pig (15 %), rabbit (10 %), mouse (5 %), and dog (5 %) (Fig. 1b). From all 167 studies, only 4 reported laparoscopic surgery in animals, 3 in pig models [10, 37, 38], and 1 in a rat model [39]. One study performed transanal endoscopy [40], while all other studies used laparotomy. Research was mainly performed in Europe, with several research groups responsible for a relatively large contribution to the total number of articles (Fig. 1c).

Outcome measures reported in the included studies were anastomotic healing, anastomotic leakage, bursting pressure, tensile strength, adhesion scores, and histological parameters such as influx of granulocytes and collagen deposition. We summarized the characteristics of the used interventions and outcome measures that were used in these studies. This summary was sent to panel members as background information when completing the questionnaires. (PDF of the used questions during the Delphi rounds can be found online as supplementary data, S2, S3, and S4).

In total, 44 authors were willing to participate in the Delphi analysis, together being responsible for 77 of the included articles from the major research groups worldwide (Fig. 2). The first questionnaire was completed by 39/44 responders (89 % response rate). After non-responders were excluded, the second round was completed by 35/39 panel members, all of whom also completed the final round (Table 1). No additional people were invited as the Delphi progressed.

The occupation of the panel members is summarized in Table 2. Researchers had operated on approximately 200 animals and have published a median of four articles regarding this topic. All panel members were affiliated with a university and have experience performing animal research on this topic. The institutions listed in Table 1 were not involved in this project and do not necessarily subscribe to the consensus.

The first questionnaire consisted of 95 items in six categories, namely the choice of animal model, location and technique of the anastomosis, macroscopic outcome, histological assessment, mechanical and biochemical outcome measures, and reporting specifics on animal research. After the first round, consensus existed on 58 of the items (61 %) and disagreement or uncertainty on 37 items. Based on additional remarks, seven items were added and two were rephrased. The seven newly introduced items for the second round consisted of additional techniques in the category mechanical and biochemical outcome measures and a suggestion by one of the panel members for a new macroscopic scoring system.

The second questionnaire was based on the first and consisted of 37 items on which consensus did not exist in the first round and 7 items added based on additional comments (total of 44 items). The most important item that was added in round 2 and on which immediate consensus was reached was the newly introduced Anastomotic Complication Score (ACS, see Table 3), which was proposed during the first round by one of the panel members. After the second round, consensus existed only on 3 items, in addition to the 58 items on which consensus was reached in the first round.

Feedback was provided to the participants after each round. In the final round, a clear distinction was made between positive and negative arguments from the second round to simulate a discussion between panel members. The topics for debate remained why certain animal models should or should not be used. Some panel members argued that small animal models are not able to reflect the clinical setting while others are certain that with the right scoring systems, one can obtain sufficient information to make the model more translational. Based on all arguments given by the panel members and the first two rounds, 20 recommendations were proposed. Consensus was reached on 17 of those recommendations (Table 4).

In this study, consensus was reached among researchers studying intestinal anastomoses in animals. The main result is that the selection of an animal model depends on the research question and there is no “one size fits all.” Consensus was reached that mouse, rat, and pig models are considered appropriate models but dog and rabbit models should no longer be used for research on bowel anastomoses in the lower GIT.

The main outcome of the study (anastomotic healing/leakage) should always be evaluated macroscopically, where currently used scores were not considered appropriate enough. The Anastomotic Complication Score, as proposed by a panel member, may provide an objective scoring measure. Obviously, this new score needs to be evaluated in the experimental setting to obtain information about veracity and/or inter-observer variation, but it does seem to be a promising tool. Bursting pressure (or tensile strength) together with histological evaluation provides further information about the anastomosis. Additional analyses can be helpful to answer specific research questions but are not (yet) considered appropriate as surrogate markers for anastomotic healing. Reporting on animal testing and welfare is still not detailed enough in current literature and ARRIVE guidelines should be followed as much as possible.

A summary of items on which consensus was reached can be found in Table 4. A complete overview of the Delphi process with more detailed outcomes can be found online (S2, S3, and S4).

The first topic of debate was the use of small animals. Many panel members felt that they were appropriate to use when systemic interventions are tested, but when interested in a local device, a larger animal is preferred. Although it is obvious to prefer to test devices intended for human use in an animal model of comparable size, testing a local device in a rat model is also considered acceptable. Mice should not be used to answer research questions on local devices. Another topic of debate was the use of mice and rats as models for healing or leakage. Many panel members felt that rats cannot be used as a model for AL since they are more resistant to infections and show hardly any clinical signs while other panel members have been using rats for this purpose for many years with very good results. Even though the rat is a validated model for both anastomotic healing and leakage, there are still opponents that claim that a rat is not suitable for this purpose, mostly based on own experiences. There was also disagreement on the consideration of practical ease in large animal models. Some believe that they are difficult in terms of anesthesia and housing, while others find them rather easy to handle and do not see any practical disadvantages.

For clinical translation to the human setting, all panel members agreed that the pig was the best-suited model in the preclinical setting. However, the use of mice as a clinically relevant model was also suggested, because it might mimic clinical AL better than the rat model. Despite solid arguments and a clinical scoring system proposed by one of the panel members, there was no consensus on the use of small animal models for clinical translation. The proposition to first test a hypothesis in a small animal model (mouse, rat, rabbit) and then use a large animal model (pig or dog) to make it more clinically transferable was not agreed upon and is therefore not recommended.