Predicting risk factors for postoperative intestinal stenosis in neonates with necrotizing enterocolitis: development and assessment of a predictive nomogram Authors: Yang Chen, Ling Zhou, Qianghui Liao, Dong Xiao, Ledao Zhu, Jinlong Yao, Jiashu Liu, Xuling Zhang, Xiaopeng Ma, Feng Ren, Zhiyong Wang Published: 2024 ABSTRACT AbstractObjectiveThe aim of this study was to develop and validate an intestinal stenosis prediction model for postoperative newborns with neonatal necrotizing enterocolitis (NEC). MethodsClinical information was collected on neonates who had undergone anastomosis or enterostomy because of NEC. The least absolute shrinkage and selection operator regression was applied to identify risk factors included in the model for postoperative intestinal stenosis. Multivariate logistic regression analysis was used to develop a predicting model regression based on the selected variables. Then internal validation was assessed using the bootstrapping validation. The accuracy and applicability of the model are assessed by C-index, calibration and decision curve.ResultsPredictors incorporated into the model were a weight on admission, hematochezia, duration of abnormal C-reactive protein, lactate, intestinal peristalsis vanish, operation methods and duration of surgery. The regression equation was logit (P) = −0.001X1 + 1.566X2 + 0.185X3 + 0.304X4 + 1.34X5 − 2.932X6 + 0.015X7 − 3.193, where X1 was weight on admission (g), X2 was hematochezia (yes = 1, no = 0), X3 was duration of abnormal C-reactive protein (days), X4 was lactate (mmol/L), X5 was intestinal peristalsis vanish (yes = 1, no = 0), X6 was primary anastomosis (yes = 1, no = 0), X7 was duration of surgery (min). The model displayed good discrimination with a C-index of 0.879 (0.827,0.932) by random sampling for 1000 times. The calibration curve excluded the overfitting performance, and the decision curve confirmed the clinical application capacity of the model. ConclusionThis nomogram of intestinal stenosis incorporating the use of weight on admission, hematochezia, duration of abnormal C-reactive protein, lactate, intestinal peristalsis vanish, operation methods and duration of surgery could be conveniently used to facilitate the intestinal stenosis risk prediction in postoperative-NEC-patients.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00383-024-05916-5.Keywords: Intestinal stenosis, Neonatal necrotizing enterocolitis, Predictive nomogram Introduction Neonatal necrotizing enterocolitis (NEC) is a devastating intestinal disease for newborns, which can cause serious complications, even death. Recent research has shown that the overall mortality rate of newborns diagnosed with NEC was 23.5%, of which 25–50% required surgical intervention [1]. With the improvement of early diagnosis and timely surgical intervention (intestinal resection and anastomosis or enterostomy), the mortality rate of NEC has decreased. However, postoperative complications, such as intestinal stenosis, adhesive intestinal obstruction, and short bowel syndrome, remained high [2]. Postoperative intestinal stenosis mainly manifests as scarring narrowing or atresia of the intestinal canal, resulting in intestinal obstruction, often requiring reoperation [3, 4]. However, no research specifically addresses the risk factors for developing intestinal stenosis after NEC surgery [5]. In this retrospective study, we collected data from children undergoing NEC surgery at a well-established children’s hospital. We developed a clinical prediction model for postoperative intestinal stenosis by exploring the possible risks and intervention factors of intestinal stenosis, so as to provide patients with better prognosis consultation and more timely clinical intervention. Information and methods Patients This study was approved by the Ethics Committee of Shenzhen Children’s Hospital (No.202205902). Clinical information of 160 patients was collected from January 1, 2016 to June 1, 2023. All patients underwent surgical treatment and were confirmed as NEC. Cases with other systemic diseases, multiple system anomalies, severe complications during treatment, abandonment or death, and follow-up less than three months after discharge were excluded. Those with intestinal stenosis identified by barium enema or surgical exploration were classified as the stenosis group, while the rest were in the non-stenosis group. Stenosis assessment The primary surgical treatment for NEC includes intestinal resection and anastomosis or enterostomy. The clinical manifestations of children with intestinal stenosis after primary anastomosis are repeated feeding intolerance, such as abdominal distension, biliary vomiting, gastric retention, and constipation. Abdominal radiographs were performed when the patient presented symptoms of obstruction. Patients with suspected stenosis were given a barium enema to determinate the location of the stenosis. Furthermore, an intraoperative diagnosis was made for severe obstruction. As for patients with enterostomy, intestinal stenosis usually occurs at the distal intestine to the stoma. The intestinal stenosis was diagnosed by a barium enema of the distal intestine to the stoma before stoma closure or intraoperative exploration [6, 7]. Variables selection To adequately explore possible influencing factors, we collected as much clinical information as possible. Firstly, as basic information, gender, age, weight, birth and admission information were naturally included. Then, on the basis of previous studies, important laboratory and imaging findings also were contained in the screening variables. Besides, indicators related to the operation were used to predict the severity of the disease and recovery from intestinal necrosis. And the effects of treatment were ultimately mainly judged by prognostic indicators. Statistical analysis R software (version 3.1.1) was applied for statistical analysis. Mean ± standard deviation was used for data conforming to a normal distribution. Otherwise, median ± quartile was used, and classification variables were expressed by percentage. Then, the least absolute shrinkage and selection operator (LASSO) regression analysis was performed to converge the predictive variables, and appropriate parameters were determined to obtain the optimal solution of variable screening. According to LASSO regression preliminary screening results, multivariable logistic regression analysis was performed to ascertain the high-risk factors affecting intestinal stenosis. The prediction equation and nomogram were constructed. The model was verified internally by calculating C-index with 1000 random samples. Moreover, the calibration curve was drawn to compare the consistency of the prediction probability and the actual probability. Finally, the decision curve describing net benefit was plotted to determine whether the predicted results benefited the clinical work. Results Patient’s information A total of 160 cases of NEC after surgery were included in the study, including 105 cases in the non-stenosis group and 55 cases in the stenosis group. The clinical information of the two groups is shown in Table 1. Baseline information such as sex and age had no statistical significance (P > 0.05). Table 1Comparison of clinical information between the two groupsVariablesNon-stenosis group (n = 105)Stenosis group (n = 55)χ2/tPSex (female)53 (50.5%)32 (58.2%)0.8610.354Gestational age/d243.14 ± 26.72233.20 ± 27.362.2170.028Age of onset/d11.28 ± 8.5311.91 ± 9.45−0.4290.668Cesarean65 (61.9%)28 (50.9%)1.7930.181Birth weight/g2298.26 ± 809.291880.96 ± 750.323.1750.002Weight on admission/g2404.57 ± 762.972011.82 ± 677.803.2100.002Ventosity96 (91.4%)52 (94.5%)0.1560.693Vomiting45 (42.9%)15 (27.3%)3.7400.053Hematochezia55 (52.4%)42 (76.4%)8.6970.003Poor reaction41 (39.0%)28 (50.9%)2.0700.150Abdominal erythema16 (15.2%)10 (18.2%)0.2300.632Abdominal muscle tension71 (67.6%)42 (76.4%)1.3300.249Borborygmus disappearance42 (40%)27 (49.1%)1.2160.270Temperature/℃37.04 ± 0.6237.13 ± 0.75−0.8200.414PH in arterial blood7.37 ± 0.097.33 ± 0.122.3450.02Leukocytes × 1099.45 ± 5.528.12 ± 5.011.4860.139Neutrophils × 1096.10 ± 4.215.32 ± 3.991.1360.258Eosinophils × 1090.40 ± 0.350.40 ± 0.47−0.0040.997Basophils × 1090.05 ± 0.050.06 ± 0.16−0.8170.415Hemoglobin/g/L118.48 ± 23.63116.96 ± 26.660.3680.714platelet × 109210.93 ± 105.47187.49 ± 101.471.3530.178C-reaction protein/mg/L56.71 ± 48.0274.99 ± 54.57−2.1800.031Duration of abnormal CRP/d6.27 ± 3.4010.40 ± 7.12−4.067 < 0.001Albumin/g/dl27.35 ± 4.6827.27 ± 4.110.1110.911Lactate/mmol/L2.39 ± 1.643.65 ± 2.06−4.203 < 0.001Procalcitonin/ng/ml23.01 ± 26.6043.50 ± 56.56−2.5440.013APTT/s53.97 ± 17.9054.97 ± 16.06−0.3470.729PT/s16.73 ± 4.9618.03 ± 6.72−1.3940.165Fibrinogen/g/L3.18 ± 1.113.21 ± 0.97−0.1520.879Intestinal peristalsis vanish19 (18.1%)27 (49.1%)16.928 < 0.001Seroperitoneum/cm1.48 ± 0.691.47 ± 0.670.0540.957Intestinal wall pneumatosis0.28 ± 0.060.29 ± 0.08−0.4690.640Primary anastomosis14 (13.3%)2 (3.6%)3.7710.052duration of surgery/min111.75 ± 34.08134.58 ± 40.73−3.759 < 0.001APTT activated partial thromboplastin time, PT prothrombin time Comparison of clinical information between the two groups APTT activated partial thromboplastin time, PT prothrombin time Variables filtration The two groups were preliminarily screened by LASSO regression analysis to draw the convergence curve of variables under different parameters (Fig. 1). When the curve tended to be parallel, the applicable parameter was selected to obtain the optimal solution by five times cross-validation (Fig. 2). The final preliminary screening variables were weight on admission, hematochezia, duration of abnormal C-reactive protein, procalcitonin, lactate, intestinal peristalsis vanish, operation methods and duration of surgery. All variables were preliminarily screened by lasso regression The optimal solution was obtained through 5 times of cross certification Model construction The filtered variables were included in the multivariable logistic regression analysis, indicating that procalcitonin was eventually excluded. Detailed information on each risk factor is shown in Table 2. The regression equation was logit (P) = −0.001X1 + 1.566X2 + 0.185X3 + 0.304X4 + 1.34X5 − 2.932X6 + 0.015X7 − 3.193, where X1 was weight on admission (g), X2 was hematochezia (yes = 1, no = 0), X3 was duration of abnormal C-reactive protein (days), X4 was lactate (mmol/L), X5 was intestinal peristalsis vanish (yes = 1, no = 0), X6 was primary anastomosis (yes = 1, no = 0), X7 was duration of surgery (min). Then, according to the above result, the nomogram was plotted to represent the prediction model (Fig. 3). Table 2Multivariate regression analysis for complete enteral nutrition after intestinal atresiaVariablesBWaldPExp (B) (95% Cl)Weight on admission−0.00113.149 < 0.0010.999 (0.998–0.999)Hematochezia1.5669.4750.0024.785 (1.766–12.967)Duration of abnormal CRP0.1859.1160.0030.053 (0.007–0.416)Lactate0.3045.5050.0191.015 (1.003–1.027)Intestinal peristalsis vanished1.3407.3830.0071.203 (1.067–1.357)Primary anastomosis−2.9327.8130.0051.356 (1.051–1.748)Duration of surgery0.0156.0310.0143.817 (1.453–10.031)Constant−3.1927.8250.005– Multivariate regression analysis for complete enteral nutrition after intestinal atresia The calibration curves of the nomogram for complete enteral nutrition Model evaluation C-index was calculated as 0.879 (0.827, 0.932) by random sampling 1000 times, implying that the model had high accuracy. The calibration curve (Fig. 4) suggested a good fit between the predicted and actual probability. In addition, the Hosmer–Lemeshow test showed that the Chi-square value was 8.601, P = 0.377 > 0.05, which revealed no overfitting performance in the prediction model. Besides, the decision curve describing the net benefit (Fig. 5) showed that the benefit curve was located at the upper right, illustrating that the clinician’s judgment based on the prediction model can primarily benefit the patient. The calibration curve for estimating overfitting performance of the prediction model The decision curve for describing the net benefit of the prediction model Discussion The currently accepted pathogenesis of intestinal stenosis is the injury to the muscular layer caused by ischemia or infection. Preservation of suspected necrotic bowel is an important cause of postoperative intestinal stenosis [8]. However, in clinical practice we have found that in children with NEC undergoing enterostomy surgery, stenosis is rarely observed in the proximal end of the stoma, whereas the distal end of the stoma is a frequent site of intestinal stenosis, which suggests that the presence of intestinal stenosis may be related to several factors, including the initial surgical approach. Risk assessment of intestinal stenosis allows pediatricians to predict the likelihood of its occurrence in combination with a number of factors, and thereby reduce the likelihood of postoperative reoperation for stenosis. Previous studies on intestinal strictures have not been analyzed for postoperative strictures. Therefore, we have developed a comprehensive assessment system for postoperative intestinal stenosis by means of objective indicators. According to the results of statistical analysis, the variables finally determined to be included in the model were weight on admission, hematochezia, duration of abnormal C-reactive protein, lactate, intestinal peristalsis vanish, operation methods and duration of surgery. The children’s weight on admission was a protective factor for intestinal stenosis. The weight was affected by gestational age and onset time, reflecting the nutritional status of children. Lower body weight is associated with a greater risk of more severe NEC, which also increases the probability of complications [9]. Hematochezia was an intuitive manifestation of intestinal mucosal damage, and it was more objective and realistic than other clinical symptom descriptions. In fact, bloody stools are also important evidence for the clinical determination of NEC [10]. However, with the development of medical technology and postpartum care, current studies had shown that there was no significant relationship between the mode of delivery and the severity of NEC, while only multiparous pregnancy had been found to increase the chance of NEC [11]. In addition, as a non-specific inflammatory indicator, C-reactive protein reflects the process of tissue damage and repair, and the injury caused by surgery will also cause its increase. Therefore, individual CRP levels are unreliable, while the duration can embody the gradual resolution of inflammation in the body. Similar studies have shown that the abnormal duration of CRP in children with intestinal stenosis was significantly longer than that in the non-stenosis group, which may be due to the slow absorption of intestinal residual necrosis[12]. Elevated lactate can suggest persistent hypoxia in the intestines to some extent, as the body always prioritizes the supply of vital organs like the brain and the heart. While arterial PH is more susceptible to breathing and even artificial fault [13]. Besides, in the early stage of NEC, seroperitoneum, hemodynamic disorder, intestinal dilation, bowel thickening even pneumatosis intestinalis may appear, which are all indicators that ultrasound can detect. The complete disappearance of peristalsis indicated that the structure and function of bowels had been seriously damaged. A meta-analysis had reached the consistent conclusion that the intestinal peristalsis vanish was closely related to serious complications and death in newborns [14]. Primary anastomosis and enterostomy were the main surgical methods in the acute stage of NEC. The essential difference between the two surgery is whether the intestinal continuity is intact. The results of this study hinted that neonates who underwent enterostomy were more likely to develop intestinal stenosis than those who underwent straight anastomosis. The possible explanation was that the integrity of the intestinal tract ensured the bowels were continuously nourished and stimulated by chyme, which promotes the recovery of the damaged tissue. Furthermore, for avoiding poor healing of the anastomosis, surgeons tend to take a more aggressive approach to remove the suspected necrotic bowels during anastomosis [15, 16]. Meanwhile, the duration of surgery was not only affected by the severity of NEC disease, protracted anesthesia and sedation, as well as vasoconstriction caused by intestinal exposure also may aggravate the damage of bowels, all of which would increase the possibility of postoperative intestinal stenosis [17]. Conversely, common indicators of infection such as leukocytes, neutrophils, and procalcitonin were not included as risk factors. These results may be related to the insensitivity of the immune system in newborns, and even the level of leukocytes would decrease in severe infection. In addition, as a specific indicator of bacteria, procalcitonin was not necessarily elevated in tissue damage caused by ischemia, autoimmunity or viruses, implying that the occurrence of intestinal stenosis may not be all the cause of a bacterial infection [18, 19]. Besides, for fear of excessive intraoperative bleeding, the coagulation function of patients would be deliberately corrected, such as transfusion of plasma, platelets or coagulation factors, which may be the reason for the exclusion of coagulation function indicators [20]. Regrettably, because of the low incidence of NEC surgery and the prolonged follow-up for complications, this study could not prospectively collect new cases for external verification. Future multicentric research may be the better solution. In conclusion, this model had a satisfactory prediction effect. Clinicians starve for an objective evaluation standard to make more advisable intraoperative decisions. The results of this study may assist surgeons in taking the appropriate surgical procedures during the first operation for NEC. The high predictive risk of postoperative intestinal stricture may suggest a more significant concern for the condition of the distal bowel. Meanwhile, through the predicted results of the model, the assessment of the incidence can help the pediatrician better explain the subsequent progression to improve the parental attention and compliance. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (XLSX 78 kb) Author contributions YC and ZW together completed data collection, statistical analysis and manuscript writing. XP, XM and FR provided conception and design of the study, participated in the modification of the manuscript. Others contributed to data arrangement and article revision. Funding This work was supported by Shenzhen Fund for Guangdong Provincial High-level Clinical Key specialties, No. SZXK035. Data availability The original data of the research had been uploaded with the manuscript and was available for peer review. Declarations Conflict of interest The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Yang Chen, Email: 939882878@qq.com. Zhiyong Wang, Email: Eswingster@gmail.com. References Associated Data Supplementary Materials Supplementary file1 (XLSX 78 kb) Data Availability Statement The original data of the research had been uploaded with the manuscript and was available for peer review.