Predictive Factors for Abnormal Diagnostic Colonoscopy Findings among Pediatric Patients: A Retrospective Cohort Study

Abstract

Introduction

This study investigated predictive factors associated with abnormal diagnostic findings of pediatric colonoscopies.

Methods

This retrospective study reviewed the medical records of children who underwent diagnostic colonoscopies between January 2004 and March 2025 at the University of Tokyo Hospital. The patients were divided into two groups based on the presence or absence of abnormal colonoscopy findings, and patient backgrounds, clinical symptoms, endoscopic findings, and blood test results were compared between the groups.

Results

A total of 112 colonoscopies were performed, among which 73 cases were the first procedure for diagnostic. Abnormal macroscopic and/or histopathological findings were identified in 36 cases within the diagnostic group. In the multivariable analysis, family history of inflammatory bowel disease or hereditary polyposis (odds ratio [OR]: 6.1, 95% confidence interval [CI]: [1.13–37.7], p = 0.035), serum hemoglobin (OR 0.68, 95% CI: [0.47–0.79], p = 0.005), and serum C-reactive protein (OR 1.04, 95% CI: [1.00–1.13], p = 0.043) was independent predictors of abnormal pediatric colonoscopy findings. The predictive model using these three factors was accurate, with an area under the receiver operating characteristic (ROC) of 0.737.

Conclusion

A detailed preliminary medical history, serum hemoglobin, and serum C-reactive protein may help predict the necessity of pediatric colonoscopy, potentially improving diagnostic efficiency.

Introduction

Colonoscopies are performed worldwide to screen for colorectal cancer and in the diagnosis of various disorders affecting the small and large intestines [1]. In the field of pediatric medicine, although colonoscopy is rarely used for cancer screening, several studies have demonstrated its utility in the diagnosis of conditions such as inflammatory bowel disease (IBD), polyposis syndromes, eosinophilic gastrointestinal disorders (EGIDs), and graft-versus-host disease (GVHD) [2, 3]. In addition to endoscopic findings, histological confirmation via endoscopic biopsy is also essential to accurately diagnose these diseases [4, 5]. The number of individuals with IBD is increasing globally, as is the number of pediatric IBD cases [6]. In Japan, the prevalence of Crohn’s disease (CD) and ulcerative colitis (UC) was 4.2 and 11.0 per 100,000 population in 2004, respectively. By 2013, these figures had increased to 7.2 and 15.0 per 100,000, representing a 73.8% increase for CD and a 45.0% increase for UC [7]. Owing to the rising burden of gastrointestinal diseases among children and emphasis on obtaining early and precise diagnoses, the clinical necessity of pediatric endoscopy has steadily increased as well.

In Western countries, the use of pediatric colonoscopy is well established and guided by evidence-based protocols [8, 9]. Japan has developed similar clinical guidelines, following the example of Western countries [10]. Colonoscopy is commonly performed in children who are present with abdominal pain, hematochezia, or diarrhea, and its utilization has been increasing in recent years, supported by advances in both sedation techniques and endoscopic equipment [11]. The reported complication rate for pediatric colonoscopy in Japan ranges from 0.21 to 1%, with bleeding occurring in 0.05–0.5% of cases [12]. In contrast, a recent nationwide survey among adults in Japan reported an overall complication rate of 0.046%, with bleeding occurring at a rate of 0.008% [13], suggesting that pediatric colonoscopy carries a higher procedural risk. Additionally, unlike in Western countries, where pediatric endoscopies are often performed by pediatric specialists, in Japan, except for pediatric hospital, the decision to perform pediatric endoscopy is typically made by pediatricians, with patients then referred to gastroenterologists; the procedure itself is most often conducted by adult gastroenterologists. Consequently, both the indications for and methods of the procedure may vary depending on the institution [3]. Furthermore, none of the existing guidelines have specifically addressed the use of quantitative clinical evaluation such as laboratory test data in determining the appropriate indications of colonoscopy for pediatric patients.

This study, therefore, aimed to describe the demographics and clinical characteristics of children who underwent colonoscopy by analyzing their medical records for endoscopic and histopathological findings and investigated potential factors that could improve diagnostic accuracy.

Methods

Study Design, Setting, and Patients

Data from the hospital’s endoscopic database were used to conduct a retrospective analysis of a single-center cohort of children aged 15 years or younger who underwent colonoscopy between January 2004 and March 2025 at the University of Tokyo Hospital. Patients who underwent double-balloon endoscopy were excluded from the review. The sample size was considered adequate for the planned analyses.

Electronic medical records were reviewed to obtain patient demographic and endoscopic data. The following information was extracted from the electronic database: age (in months), sex, purpose of colonoscopy, type of anesthesia, bowel preparation regimens, family history of colonic disease, history of transplant, primary indication, abnormal macroscopic and/or histopathological findings, definitive diagnosis, endoscopic equipment used, serious complications, and laboratory findings. All bowel preparation regimens were determined by pediatricians and ensured adequate conditions for performing the examinations in all cases. A family history of colonic disease was defined as the presence of IBD or hereditary polyposis in first- or second-degree relatives. The following laboratory values were extracted from the most recent test results prior to the colonoscopy: serum albumin (Alb), serum hemoglobin (Hb), and serum C-reactive protein (CRP).

For patients who underwent multiple colonoscopies, only data from the first procedure were included, and for those who had previously undergone a colonoscopy at another facility, only the first procedure performed at our institution was included. The patients were divided into two groups based on the presence or absence of abnormal colonoscopy findings, and patient backgrounds, clinical symptoms, endoscopic findings, and blood test results were compared between the groups.

The primary outcome was colonoscopic or histopathological abnormalities in pediatric patients. Colonoscopic abnormalities included erythema, ulcers, erosions, polyps, and bleeding. Histopathological abnormalities included inflammatory changes (cryptitis, crypt abscesses, mucosal ulceration, basal plasmacytosis, crypt architectural distortion), features suggestive of GVHD (crypt apoptosis, dropout, epithelial attenuation), features of EGIDs, and other diagnostic lesions, such as cytomegalovirus inclusions or polyp histology. All biopsies were interpreted by two board-certified gastrointestinal pathologists, with discordant cases resolved by consensus. Colonoscopies were classified as normal when the procedure report indicated no abnormalities and histopathological examination results were normal.

Statistical Analysis

Data were analyzed using RStudio (version 2025.09.2; RStudio, Boston, MA, USA) with R (version 4.5.2; The R Foundation for Statistical Computing, Vienna, Austria). Complete case analysis was used due to a low proportion of missing values. Descriptive statistics were reported as percentages for categorical variables and as means ± standard deviations for continuous variables. Categorical variables were compared using the chi-squared or Fisher’s exact test, as appropriate, while continuous variables were analyzed using the Student’s t and Mann-Whitney U tests. To address small sample size and potential separation, multivariable logistic regression was conducted using Firth’s penalized likelihood approach with the logistf package. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC), and calibration was evaluated through bootstrap internal validation (1,000 resamples). Decision-curve analysis (DCA) was performed with the rmda package to assess clinical net benefit across threshold probabilities. The same dataset was used for model development and internal validation. As sensitivity analyses, we conducted age-stratified analyses to account for potential differences in the spectrum of suspected diseases across age-groups. Furthermore, because our cohort included patients with transplantation, a distinct clinical population, we applied inverse probability of treatment weighting (IPTW) to adjust for potential confounding. A p value of <0.05 was considered statistically significant.

Results

Patient Characteristics

A total of 112 colonoscopies were performed during January 2004 and March 2025 at our hospital, among which 73 cases were the first procedure for diagnostic. Cases in which colonoscopy were performed more than once were for follow-up of IBD or for therapeutic purposes in hereditary polyposis syndromes. Among the diagnostic colonoscopies, 36 patients had abnormal findings and 37 had normal findings (shown in Fig. 1).

Fig. 1.

Study flow diagram. A total of 112 colonoscopies were performed in patients younger than 15 years. Of these, 31 examinations were excluded because they represented second or subsequent procedures in the same patients. An additional 8 examinations performed for therapeutic purposes were excluded, leaving 73 first diagnostic colonoscopies for analysis. Among these, abnormal findings were identified in 36 cases, while 37 cases showed normal findings.

Of the 73 diagnostic colonoscopies, 43 (58.9%) patients were boys, with a mean age of 133.1 ± 51.19 months, and 39 (53.4%), 9 (12.3%), and 25 (34.2%) were performed with general anesthesia, intravenous sedation, and no sedation, respectively (shown in Table 1). The most common indication for colonoscopy was hematochezia (36 [49.3%]), followed by chronic diarrhea and screening (11 [15.1%] and 10 [13.7%], respectively) (shown in Table 1). The most common reported abnormalities were polyps and UC (both 7 [9.6%]), followed by nonspecific colitis and GVHD (both 5 [6.8%]), CD (4 [5.5%]), EGIDs (3 [4.1%]), cytomegalovirus enteritis (3 [4.1%]), and infectious colitis and mucosal prolapse syndrome (both 1 [1.4%]) (shown in Table 1). As the microscopic findings were nonspecific, diagnoses of GVHD and EGIDs were made based on biopsy results. In our cohort, 11 transplant recipients were included: 2 presented with hematochezia, 8 with diarrhea, and 1 underwent preoperative screening for another disease. Nine patients had undergone hematopoietic stem cell transplantation and two had received heart transplants. Among these, one hematopoietic stem cell transplant recipient undergoing screening and one hematopoietic stem cell transplant recipient presenting with diarrhea showed normal findings, whereas abnormalities were identified in the remaining nine patients. The intubation rate of the terminal ileum was 69 out of 73 cases (94.5%). No serious complications, such as bleeding or perforation, were observed in our cohort.

Table 1.

Characteristics of diagnostic colonoscopies

​	n (%) or mean±SD	Abnormal (n = 36)	Normal (n = 37)	OR (95% CI)	p value
Age, months	133.1±51.19	128.0±8.6	138.1±8.4	​	0.402
Males, n (%)	43 (58.9)	22 (61.1)	21 (56.8)	0.83 (0.33–2.12)	0.806
Body mass index	18.7±4.57	18.3±0.9	19.1±0.8	​	0.518
Type of anesthesia, n (%)
General	39 (53.4)	20 (55.6)	19 (51.3)	1.18 (0.43–3.29)	0.820
Intravenous	9 (12.3)	5 (13.9)	4 (10.8)	1.32 (0.26–7.33)	0.730
No sedation	25 (34.2)	11 (30.6)	14 (37.8)	0.73 (0.24–2.12)	0.619
History of transplant, n (%)	11 (15.0)	9 (25.0)	2 (5.4)	5.83 (1.16–29.25)	0.024
FH of colorectal disease, n (%)	7 (9.6)	6 (21.4)	1 (2.9)	9.00 (1.01–80.0)	0.039
Primary indication, n (%)
Hematochezia	36 (49.3)	22 (61.1)	14 (37.8)	2.55 (0.91–7.38)	0.059
Diarrhea	11 (15.1)	8 (22.2)	3 (8.1)	3.18 (0.68–20.41)	0.103
Screening	10 (13.7)	2 (5.5)	8 (21.6)	0.21 (0.02–1.21)	0.086
Anal fistula	7 (9.6)	3 (8.3)	4 (10.8)	0.75 (0.10–4.83)	1.000
Anemia	3 (4.1)	1 (2.7)	2 (5.4)	0.50 (0.01–10.1)	1.000
Constipation	3 (4.1)	0	3 (8.1)	0	0.238
Abdominal pain	2 (2.7)	1 (2.7)	1 (2.7)	1.02 (0.01–82.9)	1.000
Intussusception	1 (1.4)	0	1 (2.7)	0	1.000
Yield, n (%)
Normal endoscopy	37 (50.7)	​	​	​	​
Polyps	7 (9.6)	​	​	​	​
Ulcerative colitis	7 (9.6)	​	​	​	​
Nonspecific colitis	5 (6.8)	​	​	​	​
GVHD	5 (6.8)	​	​	​	​
Crohn’s disease	4 (5.5)	​	​	​	​
EGIDs	3 (4.1)	​	​	​	​
Cytomegalovirus enteritis	3 (4.1)	​	​	​	​
Infectious colitis	1 (1.4)	​	​	​	​
Mucosal prolapse syndrome	1 (1.4)	​	​	​	​
Laboratory findings
Serum Alb, g/L	​	36.7±1.3	42.9±1.4	​	0.003
Serum Hb, g/L	​	115.1±3.3	130.7±3.3	​	0.001
Serum CRP, mg/L	​	13.3±3.2	3.0±3.3	​	0.024

Bold values indicate statistically significant differences (p < 0.05).

SD, standard deviation; CI, confidence interval; OR, odds ratio; FH, family history; GVHD, graft-versus-host disease; EGIDs, eosinophilic gastrointestinal disorders; CRP, C-reactive protein.

Factors Associated with Abnormal Findings in Diagnostic Colonoscopies

We compared the clinical features and backgrounds of patients with abnormal findings against those with normal findings. In the univariable analysis, there were no differences between the abnormal and normal finding groups regarding age at the time of the colonoscopy, sex, body mass index, and indications. A history of transplant and family history of colorectal disease were much more frequent in the abnormal findings group (odds ratio [OR]: 5.83 [1.16–29.25], p = 0.024, OR: 9.00 [1.01–80.0], p = 0.039, respectively) (shown in Table 1). The levels of serum Alb (p = 0.003) and Hb (p = 0.001) were lower, while serum CRP (p = 0.024) was higher, in the abnormal findings group (shown in Table 1).

Development and Internal Validation of a Prognostic Prediction Model

Considering the ease of obtaining information during outpatient visits and the frequency with which laboratory tests are performed, we selected family history, Hb, and CRP as predictor variables, with the presence or absence of abnormal endoscopic findings as the outcome. Given the small sample size, we performed Firth’s logistic regression. The results showed that family history (OR 6.1, 95% confidence interval [CI]: [1.13–37.7], p = 0.035), Hb (OR 0.68, 95% CI: [0.47–0.79], p = 0.005), and CRP (OR 1.04, 95% CI: [1.00–1.13], p = 0.043) were significant predictors (shown in Table 2). The likelihood ratio test yielded p = 0.002. Bootstrap validation (1,000 repetitions) demonstrated an AUC of 0.737 (shown in Fig. 2), with a Brier score of 0.211. DCA indicated clinical usefulness within the threshold probability range of under 0.30 (shown in Fig. 3; Table 3). The regression equation using continuous variables was as follows: Logit(P) = 4.0149 + 1.8084 × (Family history 0 or 1) − 0.3817 × Hb (g/L) + 0.04109 × CRP (mg/L). Upon further evaluation, in the age-stratified analysis, the model demonstrated consistent discrimination across age-groups (<10 years: AUC 0.775, ≥10 years: AUC 0.756). The difference in AUCs between the two age-groups was assessed using DeLong’s test, which showed a statistically no statistically significant difference (p = 0.552). After IPTW, covariate balance between transplant and non-transplant cases was substantially improved, particularly for sex (standardized mean difference reduced from 0.22 to 0.04). Although a small residual imbalance in age remained, the effective sample sizes were preserved without evidence of extreme weights. In the IPTW-weighted logistic regression model, family history (p = 0.048), Hb (p = 0.007), and CRP (p = 0.044) remained independently associated with abnormal findings. The IPTW-adjusted model demonstrated acceptable discriminative performance, with an AUC of 0.73 (95% CI: 0.59–0.86).

Table 2.

Comparison between normal and abnormal colonoscopy for diagnosis by Firth’s penalized logistic regression

​	Abnormal (n = 36)	Normal (n = 37)	OR	p value
FH of colorectal disease	6 (21.4)	1 (2.9)	6.1 (1.13–37.7)	0.035
Laboratory findings
Serum Hb, g/L	115.1±3.3	130.7±3.3	0.68 (0.47–0.79)	0.005
Serum CRP, mg/L	13.3±3.2	3.0±3.3	1.04 (1.00–1.13)	0.043

p values with statistical significance are shown in bold.

OR, odds ratio; FH, family history; Hb, hemoglobin; CRP, C-reactive protein.

Fig. 2.

ROC curve based on the Firth’s logistic regression model showing the area under the receiver operating characteristic curve (AUC) with the 95% confidence interval.

Fig. 3.

DCA based on the Firth’s logistic regression model. It illustrating the net benefit across a range of threshold probabilities, compared with the strategies of treating all patients and treating none.

Table 3.

Cutoff-wise performance table

Cutoff (predicted probability)	Sensitivity	Specificity	LR+	LR−	PPV	NPV
0.2	1	0.1	1.11	0	0.5	1
0.3	0.815	0.333	1.22	0.56	0.524	0.667
0.4	0.63	0.767	2.7	0.48	0.708	0.697
0.462	0.63	0.833	3.78	0.44	0.773	0.714
0.5	0.593	0.833	3.56	0.49	0.762	0.694

LR, likelihood ratio; PPV, positive predictive value; NPV, negative predictive value.

Discussion

The results of our study suggest that in addition to existing indications, a family history of IBD or hereditary polyposis, low levels of serum Hb, and high levels of serum CRP may help refine the selection criteria for pediatric colonoscopy. In previous studies, the reported diagnostic findings included UC in 27%, CD in 14%, nonspecific enteritis in 10%, polyps in 6%, EGIDs in 2%, and no abnormalities in 33% of cases [3]. Another study reported UC in 33%, CD in 11%, nonspecific enteritis in 11%, polyps in 7%, and normal findings in 30% of cases [10]. A third report demonstrated IBD in 21%, nonspecific enteritis in 5%, polyps in 21%, and normal findings in 42% of cases [14]. In comparison, our institution’s data showed a lower proportion of IBD (11%), and higher rates of GVHD (7%), and normal findings (51%). The lower frequency of IBD in our cohort may be due to early referral of suspected cases to IBD specialized centers while the elevated rate of GVHD be explained by the fact that our hospital is a leading transplant center in Japan. Additionally, the high proportion of normal findings may reflect the inclusion of many patients undergoing screening colonoscopies prior to transplantation or other surgical procedures.

Common indications for colonoscopy in children include chronic or profuse diarrhea, suspected lower gastrointestinal bleeding, unexplained anemia, polyposis syndrome, and failure to thrive or uncontrolled weight loss [8, 9]. Previous reports have also indicated that hematochezia, diarrhea, and abdominal pain are common indications for endoscopic examinations in pediatric patients [3]. Similarly, in our study, over half of the colonoscopies performed to evaluate hematochezia or diarrhea. Although hematochezia did not reach statistical significance, it showed a tendency to be associated with abnormal findings, whereas isolated symptoms, such as anal fistula, anemia, or abdominal pain were not appropriate indications. Preoperative screening tended to yield normal findings, further emphasizing that the indication for pediatric colonoscopy should be determined only after careful and comprehensive clinical assessment (shown in Table 1). Therefore, rather than relying on a single indication, it is necessary to determine the appropriateness of colonoscopy comprehensively.

To the best of our knowledge, although anemia is included among the clinical indications for pediatric colonoscopies, no previous studies have proposed specific thresholds based on laboratory data associated with IBD and GVHD, such as Alb, Hb, or CRP for predicting abnormal colonoscopic findings in pediatric patients. These thresholds were found to be statistically significant and may potentially serve as useful indicators for predicting abnormal colonoscopic findings.

A significant difference was also noted regarding family history, which plays a key role in the diagnosis of hereditary polyposis syndromes and IBD. Although colorectal cancer is rare in pediatric populations, early surveillance is recommended for patients who have a family history of polyposis syndromes [15]. In our study, two asymptomatic patients with a family history of familial adenomatous polyposis were found to have multiple polyps. A report from Korea noted that 22.4% (17/76) of polyps observed on colonoscopies of pediatric patients were suspected to be premalignant [16]. These findings underscore the importance of obtaining a detailed family history whenever possible. GVHD is a major cause of morbidity and mortality in pediatric transplant recipients; however, as its symptoms are often nonspecific, obtaining a histological diagnosis through colonoscopy is essential [17]. Several pediatric colonoscopy studies have shown that the diagnostic yield varies substantially by indication, with relatively low yields in children presenting with isolated abdominal pain or preoperative screening and higher yields in those with hematochezia or abnormal laboratory tests. In contrast, adult cohorts with lower gastrointestinal bleeding often report very high diagnostic yields, frequently exceeding 90%, highlighting important differences in disease spectrum between adults and children. It has been reported that, among 390 pediatric colonoscopies performed, 20 of the 172 cases assessed as normal by the endoscopist were later found to have pathological abnormalities. 8 of the 20 patients had a known diagnosis of IBD, 4 had symptoms highly suspicious for IBD, and another 3 patients were immunosuppressed secondary to either bone marrow transplant or nephrotic syndrome [18]. It is important to recognize that, even when endoscopic findings appear normal, biopsy remains essential in pediatric patients, especially when IBD or immunological disorders are suspected.

In the present analysis, a regression model incorporating family history, decreased Hb levels, and elevated CRP was developed. Differences observed between ROC-based cutoff values and DCA likely reflect the distinct conceptual frameworks of these evaluation methods. In this clinical setting, relatively low threshold probabilities may be more relevant for decision-making. Within this range, particularly below a threshold probability of 0.30, the model showed a consistent net benefit across analyses, suggesting a possible role in supporting clinical judgment rather than serving as a definitive decision rule.

Our study had some limitations worth discussing. First, this was a retrospective, single-center study, with potential selection bias in determining the indications for colonoscopy, which were performed at the discretion of the attending physicians. Second, the sample size and the inclusion on GVHD cases – a condition specific to transplant centers – may limit the generalizing of our findings to the broader pediatric population. As previously reported, gut GVHD predominantly occurs following allogeneic hematopoietic stem cell transplantation [19], and our results did not contradict these established observations. Also, external or temporal validation could not be performed in the present study because of the limited sample size and single-center design. We plan to evaluate the performance and generalizability of the proposed prediction model using newly accumulated data in future studies, including temporal and external validation cohorts. Third, although the usefulness of fecal calprotectin had been recognized internationally since the 2010s [20], routine testing in Japan remained limited because insurance coverage was only introduced for UC in 2017 and for CD in 2022. At our institution as well, fecal calprotectin was measured in only 3 of 42 cases since 2017. All measured cases showed elevated levels and endoscopic abnormalities. Because of the large amount of missing data, fecal calprotectin was not included as a variable in this study; however, it is considered a potentially valuable marker, and further data accumulation is warranted.

Acknowledgments

We would like to thank Editage (www.editage.jp) for English language editing.

Statement of Ethics

This study was performed in accordance with the ethical principles associated with Declaration of Helsinki update 2013. Opt-out informed consent protocol was used for use or collection of participant data for research purposes. This consent procedure was reviewed and approved by the Institutional Review Board of the University of Tokyo Hospital (registration no. 11649-(3); approval date: September 7, 2017).

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This study was not supported by any sponsor or funder.

Author Contributions

Kazuki Honjo: conceptualization, data collection, data analysis, and writing – original draft. Keita Murakami: conceptualization, data collection, data analysis, and writing – review and editing. Junya Arai: conceptualization, data collection, data analysis, and writing – review and editing. Sozaburo Ihara: supervision, conceptualization, data analysis, and writing – review and editing. Hiroyuki Hisada, Ken Kurokawa, Yu Miyakawa, Mayo Tsuboi, Yuko Miura, Daisuke Ohki, Hiroya Mizutani, Chihiro Takeuchi, Seiichi Yakabi, Nobumi Suzuki, Hiroto Kinoshita, Yoku Hayakawa, Naomi Kakushima, Nobutake Yamamichi, Yosuke Tsuji, and Dai Kubota: data collection and writing – review and editing. Mitsuhiro Fujishiro: senior supervision and writing – review and editing.

Funding Statement

This study was not supported by any sponsor or funder.

Data Availability Statement

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from the corresponding author S.I. upon request.