A Diagnostic Score for Acute Small Bowel Obstruction

Background/Aim: The diagnosis of acute small bowel obstruction (ASBO) may be difficult and the decision to operate is based on clinical findings. So far, the diagnostic scores (DSs) for ASBO detection have been rarely evaluated. Patients and Methods: A cohort of 1,333 acute abdominal pain (AAP) patients with 54 ASBO patients, were included in the study. The most significant diagnostic findings (in multivariate logistic regression analysis) were used to construct DS formulas for ASBO diagnosis with location of pain at diagnosis (LP+) and without location of pain at diagnosis (LP–). Meta-analytical techniques were used to calculate the summary sensitivity (Se) and specificity (Sp) estimates for each data sets (history-taking, findings, and DS formulas). Results: In SROC analysis, the AUC values for i) clinical history-taking, ii) diagnostic findings and tests, iii) DSLP– and iv) DSLP+ were as follows: i) AUC=0.638 (95%CI=0.600-0.676); ii) AUC=0.694 (95%CI=0.630-0.724), iii) AUC=0.962 (95%CI=0.940-0.986), and for iv) AUC=0.971 (95%CI=0.952-0.988). In roccomp analysis for the AUC values, the differences are significant as follows: between i) and ii) p=0.312; between i) and iii) p<0.0001; between i) and iv) p<0.0001; between ii) and iii) p<0.0001; between ii) and iv) p<0.0001; and between iii) and iv) p=0.317. Conclusion: The present study is the first to provide data that the DS could be used for clinical diagnosis of ASBO without radiological or laboratory analyses, to reach a high diagnostic accuracy in AAP patients.

Although the diagnosis of acute small bowel obstruction (ASBO) is sometimes difficult, the diagnosis and treatment of ASBO is necessary for intestinal preservation and avoiding the risk of complications. The clinical findings of ASBO include signs of peritoneal irritation and the acute abdominal pain (AAP) with distension, abnormal bowel sounds and a history of abdominal surgery, because adhesions are complications of abdominal surgery and the most frequent etiology of ASBO (1,2). However, it is difficult to diagnose ASBO based on a single history-taking variable or a clinical finding. Physical examination of the patient is thus important for making the decision to operate or not. In practice, the decision of operation in patients with ASBO is based on doctor's clinical judgement. The lack of commonly accepted guidelines and diagnostic scoring (DS) for ASBO (1) encouraged us to evaluate the accuracy of the clinical diagnosis of ASBO among AAP patients. We designed the present study to assess the relative accuracy of i) a clinical history-taking, ii) clinical findings, as well as iii) the DS without location of pain at diagnosis (LP-) and iv) the DS with location of pain at diagnosis (LP+) in confirming ASBO among the patients with AAP.

Patients and Methods
A cohort of 1333 acute abdominal pain (AAP) patients with 54 ASBO patients was Included in the present. The clinical symptoms (n=22), signs (n=14) and laboratory tests (n=3) were recorded for each patient. The diagnosis of ASBO was confirmed by considering all clinical history-taking details, clinical findings and results of the laboratory tests and following the diagnostic criteria of ASBO as previously described (3)(4)(5)(6)(7) (Tables I and II).
Identifying the DS models. A multivariate logistic (stepwise) regression analysis (SPSS Statistics 26.0.0.1; IBM, Armonk, NY, USA) was used to disclose the variables with an independent predictive value. All the variables of clinical history and diagnostic findings presented in Tables  I and II were included in the analysis as binary data e.g. ASBO=1 and other diagnosis of AAP=0. Using the coefficients of the regression model, a DS was built and its predictive value for ASBO was studied. The coefficient of the multivariate analysis shows the relative risk (RR=en, n=ß) of a patient with a given history-taking detail, clinical finding or test to have an ASBO.
Statistical analysis. The other statistical analyses were performed using STATA/SE version 16.1 (StataCorp, College Station, TX, USA). Statistical tests presented were two-sided, and p-value <0.05 was considered statistically significant. Using 2×2 tables, we calculated sensitivity (Se) and specificity (Sp) with 95% confidence intervals (95%CI) for each clinical history-taking detail, finding or test, and using a meta-analytical technique (metaprop) separate forest plots for Se and Sp were created for each set of data, including each diagnostic variable (as equivalent to "study ID"). We calculated the summary estimates of Se and Sp, positive (LR+) and negative likelihood ratio (LR-) and diagnostic odds ratio (DOR), using a random effect bivariate model and fitted the summary hierarchical receiving operating characteristic (HSROC) curves, including all diagnostic variables in the DS LP-and DS LP+ models, using the ASBO endpoint.

The DS without location of pain (DS LP-) in ASBO.
The most important predictors of ASBO without LP at diagnosis were type of pain, previous abdominal surgery, distension and bowel sounds. The best diagnostic level for the DS LP-formula (DS V; Se=93%, Sp=95%) was reached when the patients with DS LP-values between -3.85 and -1.60, were considered as "undefined" patients for whom follow-up is required (n=188) ( Figure 5). The DS model was tested at five different cut-off levels to disclose the best diagnostic performance ( Figure 5). The Se and Sp of these five DS LP-formulas were 89% (95%CI=84-94%) and 91% (95%CI=85-96%), respectively ( Figures 5 and 6). Three of these formulas showed Se >89% and three formulas had Sp >91%. The best diagnostic DS LP-formula in these ASBO patients (formula DS V, Figures 5 and 6) showed Se of 93% (95%CI=82-99%) and Sp of 95% (95%CI=94-96%).
The DS with location of pain (DS LP+ ) in ASBO. The significant independent predictors (LP at diagnosis, type of pain, previous abdominal surgery, distension, bowel sounds) were used to build up the five different DS LP+ models. The Se and Sp of these five DS LP+ models were 92% (95%CI=88-96%) and 92% (95%CI=88-95%), respectively (Figures 7 and 8). Three formulas showed Se >92% and three formulas Sp over 92%. The DS LP+ (formula DS V,
HSROC and comparison of the AUC values. STATA (metandiplot) was used to draw the HSROC curves to visualise the pooled overall diagnostic performance of the history-taking, findings and different DS formulas in detecting ASBO (Figures 9, 10, 11 and 12). In SROC analysis, the AUC values for i) clinical history-taking, ii) diagnostic findings and tests, iii) DS LP-and iv) DS LP+ were as follows: i) AUC=0.638 (95%CI=0.600-0.676); ii) AUC=0.694 (95%CI=0.630-0.724), iii) AUC=0.962 (95%CI=0.940-0.986), and for iv) AUC=0.971 (95%CI=0.952-0.988). In roccomp analysis for the AUC values, the differences were significant as follows: between i) and ii) p=0.312; between i) and iii) p<0.0001; between i) and iv) p<0.0001; between ii) and iii) p<0.0001; between ii) and iv) p<0.0001; and between iii) and iv) p=0.317. and their DS predicted the need for surgery in 75% of the patients. However, no ROC analysis with AUC values for the ASBO patients were reported in their study.

Discussion
Schwenter et al. (9) applied a logistic regression model to construct a DS that would predict ASBO patients. Six variables correlated with ASBO and were given one point each: history of pain, guarding, C-reactive protein level ≥75 mg/l, leucocyte count (LC) >10,000/mm 3 , ascites volume ≥500 ml on CT and reduction of CT small bowel wall contrast enhancement. A DS≥3 had Se of 67.7% and Sp 90.8%, with the AUC of 0.87 (95%CI=0.79-0.95). Zielinski et al. (10) reviewed retrospectively 100 ASBO patients who had undergone CT. The statistically significant predictors of ASBO in multivariate analysis were mesenteric edema, ascites and lacking of "small bowel feces sign" in CT imaging. Their clinicoradiological DS had Se of 96% and a positive predictive value (PPV) of 90% (OR=16.4, 95%CI=3.6-75.4) for ASBO. However, no ROC analysis was performed for the ASBO patients.
Huang et al. (11) analyzed retrospectively a cohort of ASBO patients and in multivariate analysis, Tax ≥38.0˚C, sign of peritoneal irritation, LC >10,000 m 3 , CT signs (thickwalled small bowel ≥3 mm), and ascites were significantly associated with ASBO, with AUC for their DS model reaching 0.935. Bouassida et al. (12) conducted a retrospective study including 124 ASBO patients. In logistic regression analysis, six independent predictive factors of ASBO were identified: age, pain duration, Tax, LC and CT signs of reduced wall enhancement and segmental mesenteric fluid. The AUC value of their clinicoradiological score was 0.92. In a retrospective study of 97 patients with ASBO, Ozawa et al. (13) used multivariate analysis, serum lactate levels, ascites, and peritoneal irritation sign in CT and generated a DS, which showed 92% Se and 75% Sp for ASBO diagnosis. Unfortunately, no ROC analysis was available.
When the performance of the DS models was compared between ASBO and AA/NSAP patients, the trend was similar. The Se of the DS models in diagnosis of ASBO (92%, 95%CI=88-96%) was significantly higher than that in AA/NSAP patients (79%, 95%CI=72-85%). Although, Se and Sp usually behave reciprocally, it was a surprise to find that the Sp of the DS in ASBO patients (92%, 95%CI=88-95%) was similar to that in the AA/NSAP patients (95%, 95%CI=93-97%). ASBO patients DS LP-and DS LP+ . AUC values based on HSROC comparison test showed that the diagnostic performance of the clinical findings was slightly better than that of the clinical history-taking details only (p=0.312). However, as measured by the AUC values, the DS model was superior to both i) the clinical history-taking and ii) findings and tests.      Previous studies with a design similar to ours are scanty and only few retrospective studies with CT imaging have evaluated the applicability of the DS models in the ASBO population. Some previous studies have found a significant correlation between history of pain and ASBO (9,12), but in the present study, the AUC value of the DS LP-in ASBO patients (AUC=0.962, 95%CI=0.940-0.986) was almost identical with that in the DS LP+ patients (AUC=0.971. 95%CI=0.952-0.988).

Conclusion
As far as we know, there is paucity of DS models based on clinical findings and laboratory tests. Our results indicate that DS based on clinical findings could be an important part of the diagnostic algorithm, because the doctor's decision to operate patients with ASBO is based on clinical judgement (1). However, even with radiological and laboratory analyses (14)(15)(16), it may be difficult to reach a higher efficiency than the 97% AUC (Se/Sp balance) for the DS in ASBO. To date, the only DS study of patients with ASBO has not yet been closed and analyzed (17) and the present study is the first to provide data that the DS could be used for detection of ASBO without radiological or laboratory analyses to reach a high diagnostic accuracy in AAP patients.

Conflicts of Interest
The Authors have no conflicts of interest or financial ties to disclose in relation to this study.

Authors' Contributions
All Authors contributed to the collection and analysis of data, drafting and revising the manuscript, read and approved the final manuscript.