Development and Validation of a Clinical Nomogram for Predicting Complications From Pediatric Multiple Magnet Ingestion: A Large Retrospective Study

Abstract

INTRODUCTION:

This study aimed to develop and validate a reliable nomogram based on clinical factors to predict complications associated with pediatric multiple magnet ingestion, addressing the urgency and controversy surrounding its management.

METHODS:

Patients aged 0–18 years with multiple magnet ingestion diagnosed at the Shenzhen Children's Hospital between January 2017 and December 2023 were enrolled. Clinical data were analyzed using least absolute shrinkage and selection operator regression and multifactor logistic regression analyses to screen for risk factors. A model was constructed, and a nomogram was plotted. Model performance was evaluated and internally validated using the area under the curve (AUC), Hosmer-Lemeshow test, calibration curve, decision curve analysis, and 1,000 bootstraps. We calculated the optimal cutoff value, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of the prediction model.

RESULTS:

Of the 146 patients, 57 (39.0%) experienced complications. The nomogram included age, multiple ingestions, vomiting, abdominal pain, and abdominal tenderness. The AUC was 0.941, and the internally validated AUC was 0.930. The optimal cutoff value selected as a predictive value was 0.534, with a sensitivity of 82.5%, specificity of 93.3%, positive predictive value of 88.7%, negative predictive value of 89.3%, and accuracy of 89.0%. The Hosmer-Lemeshow test yielded a P value of 0.750. The calibration plot exhibited high consistency in prediction, and decision curve analysis showed excellent net benefits.

DISCUSSION:

Our nomogram demonstrates excellent discrimination, calibration, and clinical utility and may thus help clinicians accurately assess the risk of complications from pediatric multiple magnet ingestion.

INTRODUCTION

With economic development and improved living standards, various types of magnetic toys have become increasingly popular. As a result, the incidence of accidental magnet ingestion in children has rapidly increased worldwide (1–3). According to a survey from the National Electronic Injury Surveillance System, there were more than 23,756 cases of magnet ingestion in the United States from 2009 to 2019, with an annual increase of 6.1% (4). Notably, approximately 6,100 cases involved high-risk multiple magnet ingestion, accounting for one-fourth of the total, increasing from 350 cases per year in 2013–2016 to 797 cases per year in 2017–2019 (4).

Ingestion of multiple magnets is highly likely to compress the vascular supply of the intestinal loops owing to their mutual attraction, leading to ischemia, tissue necrosis, perforation, fistula formation, intestinal obstruction, secondary peritonitis, and even death (5). Complications have been reported in 32.5%–75% of cases of multiple magnet ingestion, accounting for 80% of all complications of pediatric foreign body ingestion (3,6). In most cases, a varied approach, including surgery and gastroenterology, is required for early intervention to minimize complications (5). Several management algorithms have been proposed to diagnose and manage magnet ingestion based on the number and location of magnets and symptoms (5–8). However, there is currently no international consensus on them, especially in predicting complications. Therefore, an accurate and convenient risk-assessment tool to manage pediatric multiple magnet ingestion is urgently required.

Nomograms are effective and reliable statistical tools that integrate multiple factors and have been used to create simple, intuitive graphs of statistical prediction models to quantify the risk of clinical events (9,10). This study aimed to identify predictive factors for complications in pediatric multiple magnet ingestion and develop a nomogram using these factors to aid clinicians in timely decision-making regarding endoscopy or surgery.

METHODS

Participants

This study was reviewed and approved by the Ethics Committee of Shenzhen Children's Hospital (Grant No. 202200402). The requirement for informed consent was waived.

Children aged 0–18 years who experienced multiple magnet ingestion and were admitted to Shenzhen Children's Hospital between January 2017 and December 2023 were retrospectively evaluated. The inclusion criteria were (i) history of ingestion of multiple magnetic foreign bodies or suspicion of multiple magnet ingestion on abdominal radiography and (ii) multiple magnets ingested as confirmed by conservative treatment, gastroscopy, or surgery. The exclusion criterion was incomplete data.

Data collection

Patient data, including (ⅰ) general information (sex, age, symptoms, and signs); (ⅱ) clinical information (conditions of foreign body ingestion [witnessed or self-reported, multiple ingestions, and number and type of magnets] and whether hospitalized); (ⅲ) radiological information from abdominal radiographs; and (ⅳ) treatment (conservative treatment, gastroscopy, and surgery) and complications, were collected.

Definitions and grouping

Multiple magnet ingestion was defined as ingesting ≥2 magnetic foreign bodies or magnets and metallic foreign bodies.

Multiple ingestions were defined as more than one ingestion with no witnesses or self-reporting defaulting to a single ingestion. The patients were divided into 2 groups based on the outcomes: the no-complication group (i.e., gastroscopy without complications, spontaneous discharge without intervention, or after failed gastroscopy) and the complication group (i.e., presence or potential for complications). Complications included ulcers, perforation, fistula formation, bowel infarction, bowel torsion, intra-abdominal hernia, and infection. Potential complications indicated that multiple magnets were attracted to each other in different intestinal loops, compressing the bowel and likely leading to complications in sufficient time.

Sample size

In the development of multivariable prediction models, the sample size is often based on the ratio of the number of individuals with an outcome event to the number of candidate predictors (more precisely, the number of parameters), referred to as events per variable (EPV). Based on an empirical study, a rule of thumb for at least 10 EPV was proposed and is widely accepted as a means of avoiding overfitting (11). Accordingly, we could consider a maximum of 5 variables (57 outcome events/10 EPV). All the available patients were included in this study.

Statistical analysis

The Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis statement was used to report the study elements (12). All analyses were performed using the statistical software package R (http://www.R-project.org, The R Foundation) and Free Statistics software version 1.9. All statistical tests were 2-tailed, with a significance level of α = 0.05. Continuous variables were tested for normality, and normally distributed variables were described using the mean (SD). Comparisons between 2 groups were performed using 2 independent-sample t-tests; otherwise, the median (interquartile range) was used with the Wilcoxon rank-sum test. Categorical variables were described using frequencies and proportions, and differences between the 2 groups were compared using the χ² or Fisher exact test.

We used the least absolute shrinkage and selection operator (LASSO) regression, an effective high-dimensional prediction method, to identify potential predictive variables for pediatric multiple magnet ingestion complications. The optimal value of λ was determined by fivefold cross-validation. Lambda = min was chosen to determine the final candidate features for the cross-validation results. A multivariate logistic regression analysis was performed using the features selected in the LASSO regression model to identify statistically significant predictors, which were then used to construct a nomogram.

The receiver operating characteristic curve and area under the curve (AUC) were used to evaluate the discriminative ability of the nomogram, and the Hosmer-Lemeshow test was used to evaluate the calibration of the nomogram. The optimal cutoff point of the prediction nomogram was determined by maximizing the Youden index (i.e., sensitivity + specificity − 1), and its sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were calculated. The bootstrap method was used for internal validation, and 1,000 bootstrap replicates were performed on randomly selected samples from the original data. Calibration AUC values were calculated, and calibration curves were plotted to assess the predictive ability of the nomogram. Decision curve analysis was used to assess net clinical benefits.

RESULTS

General characteristics

A total of 148 patients with multiple magnet ingestion were reviewed. Two patients were excluded owing to incomplete data. Finally, 146 patients were included in the study, including 95 males (65.1%) and 51 females (34.9%). The patient selection flowchart and grouping are shown in Figure 1. The age ranged from 0.6 to 12.0 years, with a median (interquartile range) of 4.6 (3.0–6.7) years. There were 89 patients (61.0%) in the no-complication group and 57 patients (39.0%) in the complication group. The patients in the no-complication group underwent spontaneous expulsion (n = 69; 54 patients without intervention and 15 patients with failed gastroscopy) and gastroscopy success without complications (n = 20). The patients in the complication group underwent gastroscopy (n = 11), surgery (n = 35), or gastroscopy with surgery (n = 11).

Figure 1.

Flowchart of patient selection and grouping in the study.

The following complications occurred: (ⅰ) present complications (n = 51): ulcer (n = 6), perforations (n = 45 [79%] 41 of which involved the small intestine, with 41 patients [91%] having 2 or more perforations, 14 patients [34%] having 5 or more perforations, and 1 patient having a maximum of 15 perforations), fistula formation (n = 7), intestinal obstruction (n = 29), bowel infarction (n = 14, with 25 cm of the small bowel removed in the most severe patient), intestinal volvulus (n = 2), intra-abdominal hernia (n = 2), infection (n = 20, including 3 patients aged <3 years developing septicemia or infectious shock), and no death and (ⅱ) potential for complication (n = 6). Based on the gastroscopic or surgical findings, the locations of the magnets in the complication group were as follows: small intestine (n = 40), stomach (n = 24), colon (n = 19), and esophagus (n = 5). The magnets were located in multiple gastrointestinal tracts in 36 (63%) of complicated patients, The baseline clinicodemographic patient characteristics are shown in Table 1. Of the 18 characteristics (i.e., sex, age, underlying behavioral diagnosis, number of magnets, type of magnets, eyewitness or self-report, multiple ingestions, vomiting, abdominal pain, bloody stool, fever, loss of appetite, dispirited mood, less urination, abdominal tenderness, abdominal muscle tightness, active bowel sound, and abdominal distention) collected. Seven characteristics were selected on the basis of nonzero coefficients calculated by LASSO logistic regression analysis (Figure 2a,b). The selected features included vomiting, abdominal pain, abdominal tenderness, age, type and number of magnets, and multiple ingestions. These features were then included in the multivariate logistic regression analysis.

Table 1.

Characteristics of 146 pediatric patients with multiple magnet ingestion

Figure 2.

Feature selection using LASSO binary logistic regression model. (a) Log (lambda) value of 18 features in the LASSO model. A coefficient profile plot was produced against a log (lambda) sequence. (b) Parameter selection in the LASSO model uses five-fold cross-validation through minimum criterion. Partial likelihood deviation (binomial deviation) curves and logarithmic (lambda) curves are plotted. Minimum standard and 1-SE of the minimum standard are used to draw a vertical dashed line at the optimal value. Optimal lambda produces 7 nonzero coefficients. LASSO, least absolute shrinkage and selection operator.

Risk prediction nomogram development

Multivariate logistic regression analysis identified age, multiple ingestions, vomiting, abdominal pain, and abdominal tenderness as independent predictors of complications from pediatric multiple-magnet ingestion (Table 2). These independent predictors were thus incorporated to develop a predictive nomogram (Figure 3), with a higher score indicating a higher risk of complications.

Table 2.

Predictive factors for complications caused by pediatric multiple magnet ingestion

Figure 3.

Nomogram for predicting the risk of complications from pediatric multiple magnet ingestion. The first line represents the scoring scale. Corresponding scores for each predictor factor are shown in lines 2–6. The score for each predictor is determined by referencing the first line. The total score for the risk evaluation is the sum of each predictor score. To determine the likelihood of a complication resulting from pediatric multiple magnet ingestion, the score point is located on the total point line (line 7). Then, the user descends vertically to the risk of complication (line 8).

Performance and validation of the nomogram

The receiver operating characteristic curve showed that the AUC of the nomogram was 0.941 (95% confidence interval, 0.903–0.978) (Figure 4a). The optimal cutoff value selected was a predictive value of 0.534, with a sensitivity of 82.5%, specificity of 93.3%, positive predictive value of 88.7%, negative predictive value of 89.3%, and accuracy of 89.0%. Internal validation using the 1,000 bootstraps method generated an AUC of 0.930 (95% confidence interval, 0.895–0.978), indicating good predictive discriminative power.

Figure 4.

(a) ROC curve of the nomogram. The point on the curve represents the optimal cutoff value (specificity, sensitivity). The brackets next to the area under the ROC curve (AUC) represent the 95% confidence interval. (b) Calibration curve of the nomogram. The apparent curve represents the relationship between predicted and actual probabilities of clinically significant complications. The bias-corrected curve is plotted by bootstrapping using 1,000 resamples. The ideal curve is the 45° line, which indicates perfect prediction. (c) Decision curve analysis of the nomogram. Red solid lines represent the nomogram, x axis, cutoff probability, and y axis, net benefit. AUC, area under the curve; ROC, receiver operating characteristic.

In addition, the Hosmer-Lemeshow test showed a good fit (P = 0.750). In the internal validation using 1,000 bootstrap resamples, the calibration curve showed high consistency between the predicted and actual findings (Figure 4b).

Decision curve analysis assessment of the clinical utility of the nomogram showed that for all probabilities, applying the nomogram added a more significant net benefit compared with the treat-all or treat-none strategies (Figure 4c).

DISCUSSION

The ingestion of multiple magnets involves a high rate of dangerous complications requiring additional gastroscopy or surgical treatment. Our data show that more than 80% of the patients required hospitalization, more than 60% underwent gastroscopy and/or surgery, and nearly 40% experienced complications, consistent with previous studies (13,14). Complications included ulceration, perforation, fistula formation, intestinal obstruction, bowel infarction, bowel torsion, intra-abdominal hernia, and infection. The most common complication was gastrointestinal perforation (79%), similar to other studies (13,15,16). Furthermore, we found that perforations tended to occur at multiple sites: more than 90% had 2 or more perforations, and more than 30% had 5 or more perforations, with 15 perforations observed in the most severe patients. Gastrointestinal perforations caused by multiple magnets are often located in multiple places and are scattered and hidden (17). In the current study, multiple magnetic foreign bodies were dispersed in different locations throughout the gastrointestinal tracts in 63% of complicated patients. Therefore, we recommend dissection instead of laparoscopic exploration in severe cases. Dissection enables a more thorough examination of the entire gastrointestinal tract, from the stomach to the rectum, to ensure no gastrointestinal perforations or magnets are missed. It is necessary to review the intraoperative radiographs if the number of magnets is uncertain before surgery.

In addition, the incidence of gastrointestinal perforations was significantly underestimated. In the complication group, 45 patients (79%) exhibited perforations, whereas only 2 patients (3.5%) exhibited free pneumoperitoneum on abdominal radiographs, consistent with previous results (18–20). One possible explanation is that the bowel wall around the perforation remains tightly closed, and the intestinal contents do not heavily leak from the perforation (21). This underscores the clinical importance of taking a careful history and performing a physical examination rather than relying excessively on imaging to avoid delays in diagnosis and treatment, which can lead to severe consequences (22).

Valuable experience regarding the impact of symptomatic factors has been gained. However, multiple magnet ingestion in children is primarily asymptomatic in the early stages, with vomiting and abdominal pain being the most common symptoms once the gastrointestinal function is compromised (21,23). This is confirmed in our study. According to the NASPGHAN Magnetic Foreign Body Management Protocol, gastroscopic or surgical intervention is necessary once a child develops gastrointestinal symptoms (6). However, it is essential to consider whether all gastrointestinal symptoms are equally dangerous. A multicenter retrospective study from 2 tertiary care children's hospitals in the United States analyzed 60 cases of multiple magnet ingestion and found that abdominal pain was an independent risk factor for the need for surgery (24). Another study found that positive signs of abdominal muscle tension and symptoms of abdominal pain were independent risk factors for requiring surgery (13). Although previous studies have explored the association between symptoms, signs, and interventions, our study focused on the intrinsic link among symptoms, signs, and the onset of complications. We identified a complete set of independent risk factors for complications in 146 children; these included vomiting, abdominal pain, and abdominal tenderness. Although symptoms and signs such as bloody stools, fever, dispirited mood, less urination, and abdominal muscle tightness were not independent risk factors in this study, they imply the possibility of more serious intestinal tube injuries, peritonitis, and even infectious shock that require urgent resuscitation.

The management of pediatric multiple magnet ingestion remains controversial, even in asymptomatic cases (6,8). Most asymptomatic patients in this study did not experience any complications. Therefore, it is essential to consider other variables that may predict complications. This study identified multiple ingestions as a predictor of complications associated with pediatric multiple magnet ingestion. Multiple magnets are prone to scattering across different parts of the gastrointestinal tract, especially with longer intervals between ingestions (8,22). Even when initially distant, these magnets can shift positions owing to peristalsis, attracting each other and compressing the intestinal wall, which may result in complications such as ischemia or perforation. Therefore, urgent removal of magnets in cases of multiple ingestions is necessary, even in asymptomatic patients, as long as they can be retrieved endoscopically by esophagogastroduodenoscopy or colonoscopy to prevent further complications (6).

In addition, the current study found a generally lower age in the complication group than in the no-complication group (3.7 [2.2–5.6] years vs 5.5 [3.7–7.7] years). Age was a predictor of complications from pediatric multiple magnet ingestion. The risk of complications decreases with increasing age, indicating that younger individuals are more likely to experience complications. Several factors might have contributed to this observation. First, younger children are more likely to be in the oral exploratory phase, and ingestion may occur without being witnessed or self-reported. Accidental ingestion is often detected only when an abdominal radiograph is obtained at the onset of symptoms. Second, Kabre et al demonstrated that the force exerted by the magnets on each other increased by a factor of 4 when the distance between the 2 magnets was reduced by half (25). Tsai et al discovered that magnets, such as neodymium-iron-boron, attract each other in close proximity (3.5–4.6 cm) (26). Younger age is associated with a smaller abdominal cavity space; thus, multiple magnets attract each other more easily in younger patients, increasing the likelihood of complications. Furthermore, the current study found that younger patients were more likely to develop serious complications; e.g., septicemia or infectious shock occurred in 3 patients younger than 3 years. Previous reports have shown that deaths related to these complications are most common in children younger than 2 years (6,27,28). Therefore, managing multiple magnet ingestion requires closer observation and more aggressive management in younger children.

The nomogram exhibited excellent discrimination, calibration, and clinical utility, with high sensitivity, specificity, and accuracy. The higher the score, the greater the risk of complications, indicating the need for prompt gastroscopic or surgical treatment. Conversely, a lower score indicated a lower risk of complications and a higher probability of spontaneous expulsion of the magnets. In these patients, close clinical observation and serial radiography should be performed. Performing gastroscopy or surgery on these patients may be unnecessary and increase the risk of complications associated with gastroscopy, surgery, and anesthesia (29).

Limitations

This study has some limitations. First, the data were derived from a single-center cohort and retrospectively analyzed, which may have introduced a degree of confounding bias. Second, owing to the limited number of patients, some predictors from the nomogram, such as vomiting and multiple ingestions, exhibited high point estimates and considerable variability in confidence intervals. Future studies should prospectively collect more data from multiple centers to externally validate the nomogram developed in this study and improve its feasibility and generalizability.

We developed and internally validated a reliable and user-friendly clinical nomogram to evaluate the risk of complications from pediatric multiple magnet ingestion. The nomogram can benefit clinicians and patients owing to its ability to give individual predictions of complications.

CONFLICTS OF INTEREST

Guarantor of the article: Yizhen Luo, MD, and Hongwu Zeng, MD, PhD.

Specific author contributions: This study was designed by Y.L., X.C., J.Z., Z.L., Y.G., and H.Z. Y.L., C.Z., J.Z., and X.C. collected the data. Y.L., Y.Z., and Q.S. analyzed the data. Y.L. drafted the manuscript. Y.L., Y.G., Z.L., and H.Z. revised the manuscript. Finally, all authors discussed the results and contributed to the final manuscript. All authors had full access to all the study data and were ultimately responsible for the decision to submit the study for publication.

Financial support: This work was supported by the Sanming Project of Medicine in Shenzhen (SZSM202011005) from Shenzhen Medical and Health Project and by the Guangdong High-level Hospital Construction Fund (ynkt2021-zz47).

Potential competing interests: None to report.

Study Highlights.

WHAT IS KNOWN

• ✓ The incidence of pediatric multiple magnet ingestion is increasing worldwide.

• ✓ Pediatric ingestion of multiple magnets may lead to serious complications and even death.

WHAT IS NEW HERE

• ✓ A clinical nomogram for predicting complications from pediatric multiple magnet ingestion was developed and validated.

• ✓ This nomogram included 5 risk factors: age, multiple ingestions, vomiting, abdominal pain, and abdominal tenderness.