In this study, we constructed a nomogram for the prediction of complicated appendicitis. This nomogram can be used to estimate the probability of complicated appendicitis based on the results of a logistic regression analysis. The nomogram includes one CT feature and two clinical features that are routinely obtained. This model can accurately predict complicated appendicitis, and furthermore, this model can be used to guide decision-making regarding antibiotic therapy in AA.
AA is one of the most common abdominal emergencies in the world, and the cause of the condition is still unclear; in addition, there has been little progress in this field in the past few decades[2]. To stratify patients based on the necessary clinical management, AA is divided into uncomplicated and complicated appendicitis; however, many patients still have ambiguous diagnoses, which is one of the most challenging problems. Recently, several systematic reviews and meta-analyses have shown that most uncomplicated appendicitis patients can be treated with antibiotics first [7–9]. Therefore, the World Society of Emergency Surgery (WES) guidelines recommend the use of antibiotic therapy as a safe alternative to surgery for patients with uncomplicated appendicitis and without appendicolith. Notably, the guidelines also indicate the possibility of treatment failure and the misdiagnosis of complicated appendicitis[13].
A recent meta-analysis by Podda et al. showed that the failure rate of antibiotic treatment and the recurrence rate at the 1-year follow-up were 8.5% and 19.2%, respectively. However, the failure rate of antibiotic therapy in the Vons et al study was 12% (14/120). Notably, in this study, despite CT-scan assessment, 21 (18%) of 119 patients were unexpectedly identified as having complicated appendicitis during surgery [11]. Salminen et al showed that 4 (1.5%) of 237 patients were identified as having complicated appendicitis during surgery [14]. This may be due to inaccurate CT scans or the progression of uncomplicated appendicitis to perforation.
Neither CT nor emergency MRI can be used to discriminate between nonperforated and perforated appendicitis. Leeuwenburgh et al. found that both methods incorrectly classified more than half of the patients with perforated appendicitis as having uncomplicated appendicitis[10]. A systematic review and meta-analysis of 10 CT features (extraluminal appendicolith, abscess, appendiceal wall enhancement defect, extraluminal air, ileus, periappendiceal fluid collection, ascites, intraluminal air, intraluminal appendicolith and PFS) for the diagnosis of complicated appendicitis showed that PFS was the only feature that had a high sensitivity (94%; 95% CI: 86%-98%), although it had a low specificity (40%; 95% CI: 23%, 60%)[9]. Another study showed that the pooled sensitivity of the presence of any of the 10 CT features was higher than that of individual assessments (92% vs 64%; p < 0.001), although the pooled specificity was lower (43% vs 76%; p < 0.001). In our study, the AUC, sensitivity and specificity of PFS were 79.2%, 98.7% and 59.7%, respectively. However, sensitivity, rather than specificity, should be given priority in the diagnosis of complicated appendicitis. Low specificity may lead to appendectomy in some patients who could be treated with antibiotics.
To improve the predictive ability, our nomogram includes clinical features and imaging features. Clinical features, especially when two or more are combined, have been suggested as being valuable in the diagnosis of complicated appendicitis. The NLR is a simple clinical inflammatory marker, and the NLR provides information about two different immune and inflammatory pathways, which may make it a good marker for predicting appendicitis and its severity[15–17]. In our study, the optimal cutoff value for the NLR was 7, and an NLR greater than 7 indicated the presence of complicated appendicitis. Similar to other studies, the AUC, sensitivity, and specificity of the NLR for predicting complex appendicitis were 0.667, 0.735, and 0.535, respectively. In the early stage of AA, the sensitivity of CRP is low. CRP may be more sensitive for the detection of perforation of the appendix and the formation of an abscess, although the positive predictive value of traditional inflammatory markers is relatively lower[18]. A systematic review and meta-analysis demonstrated that the CRP level was more accurate (AUC = 0.75, 95% CI: 0.71 to 0.78) than the WBC count and procalcitonin level[19]. In our study, the AUC, sensitivity, and specificity of CRP for predicting complicated appendicitis were 0.741, 0.573, and 0.841, respectively.
Previously, Atema et al. constructed a model for predicting complicated appendicitis based on clinical and imaging characteristics. The model includes five clinical indicators (age, TEMP, duration of symptoms, WBC count, and CRP level) and three CT-based parameters (presence of extraluminal free air, periappendiceal fluid and appendicolith). Nonetheless, scores > 6 points are not specific for complicated appendicitis; instead, other complex diseases, such as perforated diverticulitis and Crohn's disease, may also be indicated[20]. These limitations and the time needed to calculate up to 22 points per patient may limit its daily use. In addition, the study lacked a validation group. Avanesov et al. constructed a model for predicting complicated appendicitis based on a combination of clinical and CT features called the appendicitis severity index (ASPI)[21]. A score ≥ 4 points predicted complicated appendicitis with a positive predictive value of 92% and a negative predictive value of 83%. Again, the study lacked a validation group. In our study, evaluation of the nomogram showed that a score greater than or equal to 11 points indicated the presence of complicated appendicitis, and fourteen patients (3.7%, 14/375) and seven patients (4.2%, 7/168) with complicated appendicitis were classified as having uncomplicated appendicitis in the primary and validation cohorts, respectively. The false-negative classification of complicated appendicitis patients who then receive antibiotic treatment may result in an increased rate of treatment failure. In this study, we used an independent cohort who received antibiotic therapy to test whether this nomogram risk score could guide decision-making with regard to antibiotic therapy. When the nomogram risk score was more than 11 points, the failure rate of antibiotic treatment was 49.2%; in contrast, when the nomogram risk score was less than 11 points, the failure rate of antibiotic treatment was only 5.3%, which was lower than previously reported. This suggests that the nomogram risk score not only accurately predicts complicated appendicitis but can also be used to guide the use of antibiotic treatment in patients with AA.
Several limitations of this study need to be addressed. First, this study was retrospective and included a relatively limited number of patients from only one hospital. Although the model was validated in an independent validation cohort, there was no external validation to ensure that the results support universal application. The second is the definition of complicated appendicitis in this study. The definition of complicated appendicitis is not yet clear. The definition of complicated appendicitis in this study included gangrenous and perforated appendicitis, which are determined based on postoperative pathological results. Patients with appendiceal abscesses or inflammatory masses were excluded because this subgroup of appendicitis requires special treatment.