Predictive Factors of Acute Complicated Appendicitis: A Retrospective Study

Background: Complicated appendicitis is an indication for emergency surgery. Therefore, the predictive factors for appendicitis based on the patient background needs identication. Previously, factors predicting non-complicated and complicated appendicitis were reported. However, most of those reports were deemed unsuitable as a standard for emergency use, since those comprised too many items as predictors. We previously reported three items that preoperatively predicted complicated appendicitis (body temperature, C-reactive protein, and uid retention around the appendix). In this study, we re-evaluated different cases to conrm the usefulness of these three items can for accurately predicting complicated appendicitis preoperatively. In addition, we compared the effectiveness of these predictor items with those reported by other researchers. Methods: We retrospectively evaluated 417 adult patients who underwent surgery for acute appendicitis between January 2013 and December 2019, and compared our predictor items with those used in previous reports on the preoperative prediction of complicated appendicitis (criteria A consisting of eight predictor items and criteria B consisting of seven predictor items). Results: The area under the receiver operating characteristic curve (AUC) for the sensitivity to diagnose complicated appendicitis according to our criteria, criteria A, and criteria B were 0.823, 0.839, and 0.856, respectively. The AUC of our criteria and criteria A were similar (P = 0.356); those of criteria A and B were also similar (P = 0.352). However, the AUC of criteria B was statistically higher than that of our criteria (P < 0.05). Conclusion: Diagnostic criteria B were statistically the best predictor items for characterizing complicated and uncomplicated appendicitis. However, like criteria A and B, the AUC of our criteria exceeded 0.8, and only involved three predictor items; therefore, they can be considered useful predictors.


Background
Acute appendicitis is the most common acute abdominal disease. Annually, an estimated 43,000 cases occur in Japan; US accounts for 250,000 cases, and the UK for 50,000 cases (1). The lifetime incidence was 8.6% for men and 6.7% for women (2). In general, catarrhal appendicitis and phlegmonous appendicitis are classi ed as non-complicated appendicitis, and gangrenous/perforated appendicitis is classi ed as complicated appendicitis. There is no dispute that early surgery is the principle treatment for patients diagnosed with acute appendicitis (3). Perforated appendicitis and necrotic appendicitis with a high risk of perforation require prompt emergency surgery. Conversely, catarrhal appendicitis and phlegmonous appendicitis have a low risk of perforation in a short period of time and can be treated by quasi-emergency/quasi-waiting surgery with conservative treatment under appropriate medical supervision. Conservative treatment may be su cient depending on the course. However, the preoperative distinction between non-complicated and complicated appendicitis is challenging. In addition, cases of complicated appendicitis, including perforated appendicitis and gangrenous appendicitis, may progress to acute peritonitis, a condition that necessitates emergency surgery regardless of the time of development. There is a need for clinical indicators that accurately predict complicated appendicitis for which emergency surgery is indicated.
Studies by Avanesov et al. (4) and Atema et al. (5) report factors that predict complicated appendicitis. Avanesov et al. reported seven items as predictors: age, body temperature, elapsed time from onset, appendix diameter, uid retention around the appendix, free air, and abscess. Atema et al. reported eight items as predictors: age, body temperature, elapsed time from onset, leukocyte count C-reactive protein (CRP), free air, uid retention around the appendix, and appendicolith.
These are useful preoperative predictive factors; however, they involve many predictor items and are complicated. Therefore, there may be di culties using these predictive factors in the eld of emergency medical care. We have previously reported three predictor items of complicated appendicitis, body temperature ≥ 37.4°C, CRP ≥ 4.7 mg/dL, and uid collection surrounding the appendix on computed tomography (CT), and reported on their usefulness 6 ,7 .
In this study, we con rmed the usefulness of our criteria by comparing them with other criteria for predicting complicated appendicitis preoperatively.

Methods
In this study, we retrospectively evaluated 417 patients who underwent surgery for acute appendicitis between January 2013 and December 2019 in our institution. Blood tests and CT were performed in all cases. Patients who were successfully treated with antibiotics were excluded.
Histopathologically, catarrhal appendicitis was de ned as the apparent enlargement of lymphoid follicles in the appendix mucosa, and cellulitis appendicitis was de ned as neutrophil in ltration into all layers.
Gangrenous appendicitis was de ned as neutrophil in ltration and muscle layer necrosis, and perforated appendicitis was de ned as necrosis and perforation in all the layers. Complicated appendicitis was de ned as a pathologically proven gangrenous or perforated appendix. We classi ed all cases as complicated or non-complicated appendicitis. We compared each of the physical ndings, blood test ndings, CT image ndings, and surgical results. We determined the proportions of complicated appendicitis and non-complicated appendicitis, respectively, depending on the degree to which the criteria we reported previously (body temperature ≥ 37.4°C, CRP ≥ 4.7 mg/dL, and uid collection surrounding the appendix on CT) were met.
Next, we compared our criteria with previously reported preoperative predictor items of complicated appendicitis. The predictor items reported by Avanesov et al. in 2018 were de ned as A criteria, while those reported by Atema et al. in 2015 were de ned as B criteria. All statistical analyses were carried out using the JMP® Pro 15.0.0 (SAS Institute Inc.). The sensitivity performance for diagnosing complicated appendicitis was evaluated in terms of the area under the curve of the receiver operating characteristic curve (AUROC), homogeneity, discriminatory ability, and Akaike information criterion (AIC). The AUROC was compared between the two models using the DeLong method. Homogeneity was de ned using the likelihood ratio, χ 2 , calculated using the Cox regression model. Discriminatory ability, de ned as the large differences in survival among patients in different grades within each system, was determined using the linear trend χ 2 -test. The AIC, which shows the effect of the grading systems of sensitivity to diagnose complicated appendicitis, also estimated the results of the Cox regression model in each system. A model with a lower AIC was more explanatory and informative. Statistical signi cance was set at P < 0.05.
All study participants provided informed consent, and the study design was approved by the ethics review board of our institution. The identities of the patients were protected.

Results
There were 237 cases of non-complicated appendicitis and 180 cases of complicated appendicitis. We compared the patient backgrounds (Table 1), and patients with complicated appendicitis had signi cantly higher body temperature and CRP levels than patients with non-complicated appendicitis, in addition to signi cant uid collection surrounding the appendix. Moreover, the group with complicated appendicitis required ileocecal resection, and there were many open operations, and the operation time was long. In addition, the complicated appendicitis group had signi cantly longer postoperative hospital stays (Tables 1 and 2). We applied the risk factors reported in the past (body temperature ≥ 37.4°C, CRP ≥ 4.7 mg/dL, and uid collection surrounding the appendix on CT) to the target cases. Complicated appendicitis occurred in 86% of patients who met all three criteria. Of the patients who met two factors, 69.6% had complicated appendicitis. Of the patients who met only one factor, 31.5% had complicated appendicitis. Only 7.3% of patients who did not meet any of the factors had complicated appendicitis. The AUROC for the sensitivity to diagnose complicated appendicitis according to our criteria, criteria A, and criteria B were 0.823, 0.839, and 0.856, respectively. The areas under the curve of our criteria and criteria A were similar (P = 0.356). The A and B criteria were also similar (P = 0.352). However, those of the criteria B were statistically higher than those of our criteria (P < 0.05) (Fig. 1). The discriminatory ability linear trend χ 2 calculated using the Cox regression of our criteria and criteria A and B were 128.9, 121.1, and 142.5, respectively. The likelihood ratios χ 2 calculated using the Cox regression model, which indicate homogeneity, were 146.4, 143.6, and 172.8, respectively. The AIC of each criterion was 397.0, 345.8, 369.0, respectively (Table 3). Criteria B was the best statistical indicator. Higher discriminatory ability and homogeneity, and lower Akaike information criterion (AIC) statistics associated with better performance of the grading system.

Discussion
In our hospital, we perform emergency or quasi-emergency surgery on all patients within 24 h after admission as the standard treatment for patients with a de nitive diagnosis of acute appendicitis. Complicated appendicitis, including gangrenous appendicitis and perforated appendicitis, progresses to peritonitis when it becomes severe; thus, immediate emergency surgery is required. Therefore, a method for preoperatively predicting non-complicated appendicitis and complicated appendicitis is needed. In addition to our 2016 report (6), some predictor items have been developed to distinguish non-complicated from complicated appendicitis (4,5). They were based on patient characteristics and diagnostic markers that are routinely collected in clinical practice, combined with imaging features.
In a previous retrospective study(2, 6), we proposed predictor items to discriminate between complicated and non-complicated appendicitis using clinical and radiological ndings. Our predictor items comprised two clinical (body temperature ≥ 37.4°C, and C-reactive protein ≥ 4.7 mg/dL) and one radiological nding ( uid collection surrounding the appendix). Patients with complicated appendicitis had signi cantly higher body temperature and CRP at the time of consultation than patients with non-complicated appendicitis. CT often shows uid retention around the appendix. However, there were no signi cant differences in white blood cell (WBC) scores. The average age of patients with non-complicated appendicitis was 39 yrs, and the average age of patients with complex appendicitis was 56 yrs in this survey. Patients with complicated appendicitis were found to be signi cantly older (P < 0.001). It has been reported that WBC counts are less likely to increase in the elderly compared with CRP levels.
We have previously con rmed that our three predictor items can correctly stratify complicated and noncomplicated appendicitis. As a result, we reported that these three items are suitable predictors for complicated appendicitis preoperatively. In the present retrospective study, we examined whether our three predictor items were more effective than other predictors reported by other researchers.  (4). However, both these criteria are complicated because they consist of seven and eight predictor items, respectively, that require assessment. The criteria we adopt consists of only three factors, thus making it simpler than predictor items reported in the past. They can be used universally by both experienced and inexperienced healthcare professionals. This is a major difference from previously reported predictor items.
The analysis showed that criteria B was more effective in distinguishing between complicated and noncomplicated appendicitis. Criteria A and B implemented a combined clinical and radiological score. Both criteria included age as an evaluation item. Certainly, elderly people are more likely to have severe disease. However, elderly people often have a poor onset of symptoms, and WBC counts tend to elevate later than CRP. As mentioned above, there were no signi cant differences in WBC scores. Therefore, we adopted only CRP as a criterion.
Furthermore, criteria A includes the appendix diameter. However, in the case of an appendix that has already been perforated, this is often impossible to measure because of the in uence of surrounding abscesses and uids, and it is doubtful that it is appropriate.
Criteria A and B assess free air and abscess formation, which may be seen in other diseases, such as diverticulum perforation; thus, it is doubtful whether they are appropriate as criteria. In addition, criteria A did not include the periappendiceal uid. However, these CT ndings were 99% speci c for appendiceal perforation.
In addition, criteria A and B are time consuming as they assign different points for each item.
We believe that the predictor items for acute appendicitis should be simple. Acute appendicitis is a major disease, and doctors who do not specialize in abdominal emergencies are more likely to encounter it in their daily practice. Since our predictors are simple, they may be useful universal criteria that even inexperienced practitioner can easily use. Although our predictor assessments are limited by the retrospective nature of the study, their accuracy has been validated and their usefulness demonstrated.

Conclusions
In conclusion, the three items, body temperature ≥ 37.4°C, C-reactive protein ≥ 4.7 mg/dL, and uid collection surrounding the appendix on CT, are useful in predicting cases of complicated appendicitis preoperatively. Our predictor items can be used to quickly and appropriately predict complicated appendicitis, even by those with little experience in the settings of busy emergency care.
Abbreviations AIC Akaike information criterion; AUC:Area under the receiver operating characteristic curve; AUROC:Area under the curve of the receiver operating characteristic curve; CT:Computed tomography; CRP:C-reactive protein; WBC:White blood cells

Declarations
Ethics approval and consent to participate Informed consent was obtained from all patients or their caregivers.

Consent for publication
Not applicable.
Availability of data and materials The data and materials are not available because consent for such an action was not taken from the participants.

Competing interests
There are no competing interests to declare.  Receiver operating characteristic curve for diagnosing complicated appendicitis according to Criteria A and B. Receiver operating characteristic curve with 95% for the nal adjusted regression model. AUC, area under the curve