Upper gastrointestinal (UGI) tumor is one of the causes of acute UGI bleeding.¹ According to previous studies, the mortality rate of UGI bleeding is up to 20% in the general population and up to 45% in cancer patients with overt UGI bleeding.² Half of UGI tumor bleeding cases have chronic or latent bleeding rather than acute bleeding.³ The rate of active tumor bleeding that requires urgent hemostatic interventions such as endoscopic hemostasis, transarterial embolization (TAE), and surgery is only 15% based on endoscopic examination.⁴ In many cases, tumor bleeding does not cause massive bleeding or hemodynamic instability. However, cancer-related bleeding in cancer patients is a major cause of UGI bleeding and is associated with high mortality.² Therefore, it is difficult to predict mortality of these patients and make decisions about the necessity for urgent hemostatic intervention.

Clinical guidelines for UGI bleeding recommend the use of a bleeding risk scoring system to predict patient clinical outcomes.^5-8 The Glasgow-Blatchford score (GBS) was developed to predict the need for intervention and mortality.^9,10 Previous studies have shown that GBS has a predictive power of about 90% for intervention or death.^9-11 The Rockall score and AIMS65 were created to predict the mortality after an episode of UGI bleeding by scoring with a set of parameters.^12-14 The mortality predictive power of the Rockall score and AIMS65 is about 80%. However, in previous studies evaluating the role of the bleeding risk scoring system, the percentage of patients with UGI cancer as the cause of bleeding was only 2% to 5%.¹⁵ Despite the fact that the prognosis of patients with bleeding from UGI cancers is worse than that of patients with bleeding from other causes,¹⁶ prognostic prediction models suitable for these patients have not been well studied. Thus, it is unclear whether these bleeding risk scoring systems are useful for these UGI cancer patients.¹⁷ Thus, the purpose of this study was to identify risk factors for mortality in bleeding patients with UGI cancer and to develop a predictive model based on these factors.

1. Study design and patients

This retrospective, single-center study included patients with UGI cancers who underwent esophagogastroduodenoscopy (EGD) for clinically suspected cancer bleeding from 2009 to 2021 at Seoul St. Mary’s Hospital, Seoul, Korea. Included patients were confirmed to have UGI cancer bleeding via EGD, which was performed due to symptoms such as melena, hematemesis, sudden decrease of hemoglobin with syncope, or altered mentality. Patients with hematochezia originating from the UGI tract were also included. Exclusion criteria were: patients younger than 20 years of age, those with benign ulcer bleeding at anastomosis site after curative or palliative cancer surgery, patients with follow-up loss within 30 days of bleeding, patients with UGI cancer but with small bowel bleeding, and patients who had undergone curative cancer surgery. This study was approved by the Institutional Review Board of Seoul St. Mary’s Hospital, Korea (IRB number, KC22RISI0337). The informed consent was waived.

2. Hemostasis

EGD was performed after hemodynamic resuscitation through transfusion or intravenous fluid. If non-variceal bleeding was suspected, an intravenous proton pump inhibitor was administered before EGD. UGI bleeding was defined as active bleeding such as spurting or oozing, recent hemorrhage with visible vessel or adherent clot, and blood material found in the gastric cavity. Malignant bleeding lesions were classified using the Forrest classification.^4,18,19 Although coffee grounds-colored coagulated blood was found in the gastric cavity, only conservative treatment was administered, without endoscopic hemostasis for adherent clot, black spot, or clean base lesions. Even in cases with easy touch bleeding, conservative care was performed when blood coagulated during endoscopy. Endoscopic hemostasis was performed using hemoclips and/or electrocoagulation with or without injection. If tumor bleeding could not be treated with endoscopy, the patient was referred for TAE, radiotherapy, or surgery. Stent insertion was performed when it was determined that hemostasis was possible by pressing with a stent. In the case of rebleeding, conservative care, endoscopic treatment, TAE, surgery, and radiotherapy were determined according to the EGD findings and clinical situations.

3. Outcome assessment

We collected data about patients’ medical history, laboratory test, symptom, endoscopic findings, cancer staging, pathologic report, rebleeding, transfusion, and survival status. Based on these data, GBS, Rockall score, and AIMS65 were calculated to investigate the need for intervention and clinical outcomes for tumor bleeding.^12,20 Other organ cancer was defined as cancer in an organ other than the UGI tract or metastasis from UGI tract cancer to another organ. Rebleeding was defined as clinical signs of UGI bleeding or a decrease in hemoglobin level more than 2 g/dL with stigmata of recent hemorrhage on endoscopic findings after initial hemostasis. In this study, rebleeding within 3 days after initial hemostasis was defined as early rebleeding, while rebleeding occurring after 4 days was classified as late rebleeding.

The primary outcome of the study was patient mortality within 30 days after initial endoscopic evaluation for UGI bleeding. The 30-day mortality was defined as all-cause death within 30 days after tumor bleeding. Subgroup analysis was conducted by dividing patients into three groups according to the initial treatment method: conservative care group, endoscopic treatment group, and non-endoscopic treatment group. We compared clinical characteristics and endoscopic findings between these groups. The secondary outcome was rebleeding within 30 days after the initial endoscopy.

4. Statistical analysis

Clinical characteristics are expressed as medians and ranges for continuous variables, and as percentages for categorical variables. Continuous variables were analyzed using the one-way analysis of variance. Categorical variables were analyzed with the chi-square test or linear-by-linear association according to the number of observations. Performances of the GBS, full Rockall score, and AIMS65 in predicting clinical outcomes were determined using the area under the receiver operating characteristic curves (AUROC) with 95% confidence intervals (CI). They were categorized as follows: AUROC≥0.90, excellent; 0.80≤AUROC<0.90, good; 0.70≤AUROC<0.80, fair; and AUROC<0.70, poor.²¹ In addition, to identify factors associated with 30-day mortality, multivariate analyses using the logistic regression model were performed. Statistically significant risk factors were used to develop a new scoring system. SPSS Statistics for Windows, Version 24.0 (IBM SPSS Statistics for Windows, version 24.0; IBM Corp., Armonk, NY, USA) was used for all analyses. A p-value <0.05 was considered statistically significant.

1. Characteristics of patients

From 2009 to 2021, a total of 264 patients were included. Among them, 16 patients were excluded because of benign ulcer on anastomosis site (n=9), follow-up loss (n=5), and small bowel bleeding (n=2). In addition, patients who had undergone curative operation for their cancer lesion (n=55) were also excluded from the study (Fig. 1). Finally, a total of 193 patients were included in the present study. Baseline characteristics of the 193 patients are described in Table 1. The median age was 72 years (range, 32 to 97 years). There were 132 males (68.4%). The most common symptom was melena (60.6%). The median initial hemoglobin level was 7.7 g/dL. Intravenous proton pump inhibitors were infused in 168 (87.0%) before EGD. Bleeding from primary gastric cancer occurred in 124 (64.2%). Regarding histotypes, there were 68 (54.8%) with undifferentiated adenocarcinomas, 37 (29.8%) with differentiated adenocarcinomas, 15 (12.1%) with gastric lymphomas, two (1.6%) with gastrointestinal stromal tumors, and two (1.6%) with gastric neuroendocrine carcinomas. Among 69 patients with non-gastric cancer, there were 40 (58.0%) cases of pancreatobiliary cancer, 14 (20.3%) cases of metastatic cancer, and five (7.2%) cases of esophageal cancer. When divided by cancer stage, 18 (9.3%) patients were stage I-II, 19 (9.8%) patients were stage III, 156 (80.8%) patients were stage IV.

Figure 1. Study flow. Rebleeding within 3 days or after 4 days after initial hemostasis was defined as early rebleeding or late rebleeding, respectively.
EGD, esophagogastroduodenoscopy; UGI, upper gastrointestinal; ET, endoscopic therapy.

Table 1 . Patient Demographics and Clinical Characteristics.

Factor	Total (n=193)
Age, median (range), yr	72 (32–97)
Sex, No. (%)
Male	132 (68.4)
Female	61 (31.6)
Comorbidity, No. (%)
Diabetes mellitus	50 (25.9)
Congestive heart failure	4 (2.1)
Ischemic heart disease	17 (8.8)
Pulmonary thromboembolism	4 (2.1)
End stage renal disease	8 (4.1)
Liver cirrhosis	7 (3.6)
Cerebrovascular disease	22 (11.4)
Chronic obstructive airway disease	15 (7.8)
Other organ cancer	129 (66.8)
Medication, No. (%)
Aspirin and/or clopidogrel	29 (15.0)
Heparin or warfarin	4 (2.1)
Direct oral anticoagulant	5 (2.6)
Other antiplatelet	11 (5.7)
NSAIDs	11 (5.7)
PPI (within 4 wk)	49 (25.4)
Intravenous PPI before EGD	168 (87.0)
Bleeding symptoms at presentation, No. (%)
Melena	117 (60.6)
Hematemesis	67 (34.7)
Melena + hematemesis	19 (9.8)
Hematochezia	28 (14.5)
Syncope	3 (1.6)
Altered mental status	14 (7.3)
Systolic blood pressure, median (range), mm Hg	114 (65–180)
Pulse rate, median (range), beats/min	93 (58–141)
Initial hemoglobin, median (range), g/dL	7.7 (2.7–13.0)
Cancer type, No. (%)
Primary gastric cancer	124 (64.2)
Undifferentiated adenocarcinoma	68 (54.8)
Differentiated adenocarcinoma	37 (29.8)
Lymphoma	15 (12.1)
Gastrointestinal stromal tumor	2 (1.6)
Gastric neuroendocrine tumor	2 (1.6)
Non-gastric cancer	69 (35.8)
Pancreatobiliary cancer	40 (58.0)
Metastatic cancer	14 (20.3)
Esophageal cancer	5 (7.2)
Lymphoma	3 (4.3)
Undifferentiated adenocarcinoma	1 (1.4)
Differentiated adenocarcinoma	3 (4.3)
Other type cancer	3 (4.3)
Cancer stage, No. (%)
Stage I or II	18 (9.3)
Stage III	19 (9.8)
Stage IV	156 (80.8)

NSAIDs, nonsteroidal anti-inflammatory drugs; PPI, proton pump inhibitor; EGD, esophagogastroduodenoscopy..

2. Endoscopic findings and initial hemostasis

In the Forrest classification of endoscopic evaluation, the most common type was Ib (n=96, 49.7%) (Table 2). There were nine (4.7%) spurting cases among cancer bleeding. The median tumor size was 5.0 cm (range, 0.5 to 18.0 cm). The small sized lesions appeared in the form of vascular exposure. Patients with this had melena, hematemesis, or a sudden drop of hemoglobin level with bleeding stigmata in endoscopic evaluation. Patients were treated initially with conservative care (n=108), endoscopic treatment (n=76), or non-endoscopic hemostasis (n=9). The most used endoscopic hemostatic method was electrocoagulation (n=26, 34.2%), followed by injection alone (n=15, 19.7%) and hemostatic clip (n=13, 17.1%). Combined therapy using electrocoagulation accounted for 55.3% (n=42). Non-endoscopic hemostasis such as embolization (n=8), or stent insertion (n=1) was also performed.

Table 2 . Endoscopic Findings and Initial Hemostasis.

Variable	Total (n=193)
Endoscopic finding
Forrest classification, No. (%)
Spurting hemorrhage	9 (4.7)
Oozing hemorrhage	96 (49.7)
Nonbleeding visible vessel	15 (7.8)
Adherent clot	35 (18.1)
Flat pigmentation	19 (9.8)
Clean base	19 (9.8)
Tumor size, median (range), cm	5.0 (0.5–18.0)
Initial hemostatic therapy
Conservative care, No. (%)	108 (56.0)
Endoscopic hemostasis, No. (%)	76 (39.4)
Epinephrine injection	15
Electrocoagulation	26
Hemostatic clip	13
Electrocoagulation & injection	10
Hemostatic clip & injection	6
Electrocoagulation & hemostatic clip	4
Electrocoagulation & hemostatic clip & injection	2
Non-endoscopic therapy, No. (%)	9 (4.7)
Angiography	8
Gastric adenocarcinoma	4
Pancreatobiliary cancer	2
Duodenal cancer	1
Metastatic cancer	1
Stent insertion	1
Pancreatobiliary cancer	1

3. Subgroup analysis by initial treatment modality

Patients were divided into three groups according to their initial treatment modality. Systolic blood pressure, tumor size, and Forrest classification were significantly different among groups (Supplementary Table 1). Systolic blood pressure was significantly higher in the conservative care group than in the endoscopic treatment and the non-endoscopic treatment group (117, 110, and 109 mm Hg, respectively, p=0.03). The mean tumor size was smaller in the endoscopic therapy group (4.0 cm; range, 0.5 to 16.0 cm) than that (5.0 cm; range, 1.0 to 18.0 cm) in the conservative care group or the non-endoscopic treatment group (6 cm; range, 3.0 to 8.0 cm) (p<0.01). Types of Forrest classification were also found to be different among groups (p<0.01) (Supplementary Table 1). Lesions with Forrest IIb–III cases were mainly treated with conservative treatment (n=69, 94.5%). Lesions with Forrest Ia–IIa cases were mainly treated with endoscopic therapy (n=72, 60.0%). Most lesions that underwent non-endoscopic hemostasis were Forrest Ib lesions (n=5, 55.6%).

4. Patient outcomes and associated factors

Out of the total number of patients, 45 (23.3%) died within 30 days of initial treatment (Supplementary Table 2). Among the 45 patients who died within 1 month after the hemorrhagic event, the main cause of death was disease progression. Among them, 12 patients died from uncontrolled tumor bleeding, one from infection, one from sudden extensive cerebral infarction, and one from metabolic acidosis due to acute kidney injury. Altered mental status (odds ratio [OR], 6.0; 95% CI, 1.7 to 21.0), renal failure (OR, 5.3; 95% CI, 1.1 to 25.4), rebleeding (OR, 4.8; 95% CI, 2.1 to 10.8), age older than 65 years (OR, 3.3; 95% CI, 1.3 to 8.4), and low albumin level (<3 g/dL) (OR, 2.6; 95% CI, 1.2 to 5.8) were significant factors associated with 30-day mortality on multivariate analysis (Table 3). Tumor stage did not show statistical significance in univariate analysis. Rebleeding occurred in 50 (25.9%) (Supplementary Table 1). Initial endoscopic hemostasis was successful in 74 out of 76 patients. One patient underwent thermal therapy but did not undergo additional bleeding therapy due to acute deterioration of the patients’ general condition. The other underwent hemostatic clip, which was failed. Then, TAE was performed. We further analyzed the endoscopic hemostasis method and rebleeding, but there was no relationship between the endoscopic treatment method and the rebleeding rate (p=0.986) (Supplementary Table 3).

Table 3 . Risk Factors Associated with the Mortality within 30 Days.

Variable	Univariate analysis			Multivariate analysis
	OR (95% CI)	p-value		OR (95% CI)	p-value
Sex
Female	1.0
Male	1.6 (0.7–3.4)	0.24
Age
<65 yr	1.0			1.0
≥65 yr	2.7 (1.2–6.1)	0.01		3.3 (1.3–8.4)	0.01
Systolic blood pressure
≥90 mm Hg	1.0
<90 mm Hg	1.9 (0.6–6.1)	0.26
Pulse rate
<100 beats/min	1.0			1.0
≥100 beats/min	2.3 (1.2–4.6)	0.02		2.8 (1.2–6.2)	0.01
Hemoglobin
≥8 g/dL	1.0
<8 g/dL	1.6 (1.3–5.3)	0.20
Albumin (g/dL)
≥3 g/dL	1.0			1.0
<3 g/dL	2.6 (1.3–5.3)	<0.01		2.6 (1.2–5.8)	0.02
Blood urea nitrogen
<30 mg/dL	1.0
≥30 mg/dL	1.7 (0.8–3.3)	0.14
Prothrombin time
<1.5 INR	1.0			1.0
≥1.5 INR	4.4 (1.5–12.8)	<0.01		2.9 (0.7–11.6)	0.12
Comorbidity
Hypertension	1.1 (0.5–2.1)	0.86
Diabetes mellitus	1.4 (0.7–2.9)	0.36
Ischemic heart disease	0.7 (0.2–2.5)	0.57
Congestive heart failure	3.4 (0.5–24.8)	0.23
Liver cirrhosis	0.5 (0.1–4.6)	0.57
End stage renal disease	6.0 (1.4–26.4)	0.02		5.3 (1.1–25.4)	0.04
Pulmonary embolism	3.4 (0.5–24.8)	0.23
Chronic obstructive pulmonary disease	0.2 (0.0–1.7)	0.15
Cerebrovascular disease	1.0 (0.3–2.8)	0.95
Other organ cancer	1.3 (0.6–2.7)	0.49
Symptom
Melena	0.8 (0.4–1.5)	0.43
Hematemesis	0.7 (0.3–1.5)	0.35
Melena & hematemesis	0.2 (0.0–1.3)	0.08
Hematochezia	1.7 (0.7–4.1)	0.24
Syncope	6.8 (0.6–77.2)	0.12
Altered mental status	5.1 (1.7–15.7)	<0.01		6.0 (1.7–21.0)	<0.01
Tumor size	1.1 (1.0–1.2)	0.03		1.2 (1.1–1.3)	<0.01
Rebleeding	4.1 (2.0–8.4)	<0.01		4.8 (2.1–10.8)	<0.01
Tumor stage
Curative (stage I-III)	1.0
Non-curative (stage IV)	2.1 (0.8–5.8)	0.15
Forrest classification
Clean base	1.0
Flat pigmentation	2.1 (0.2–25.5)	0.56
Adherent clot	6.2 (0.7–53.6)	0.10
Nonbleeding visible vessel	4.5 (0.4–48.5)	0.22
Oozing hemorrhage	7.0 (0.9–55.4)	0.06
Spurting hemorrhage	9.0 (0.8–103.7)	0.08

OR, odds ratio; CI, confidence interval; INR, international normalized ratio..

5. Performance of risk scoring systems in predicting outcomes

Median overall survival after initial hemostasis was 3 months (Interquartile range, 2.3 to 3.7 months). This was significantly longer in patients with no rebleeding (4 months) than in those with early rebleeding (1 month) or late rebleeding (3 months) (p<0.001, log-rank test) (Supplementary Fig. 1). When the predictive value of 30-day mortality was evaluated, AIMS65 showed better performance (area under the curve [AUC]=0.69) than GBS (AUC=0.61) and Rockall score (AUC=0.64) (all p<0.05) (Fig. 2). We developed a new model incorporating these risk factors of 30-day mortality using logistic regression: Newscore_mortality = (if altered mental status × 1.798) + (if renal failure × 1.673) + (if rebleeding × 1.563) + (if older than 65 years × 1.190) + (if low albumin level × 0.951) – 3.363. This scoring system had an AUC of 0.79 (95% CI, 0.72 to 0.86; p<0.001) in predicting 30-day mortality. Comparisons revealed its significant superiority to GBS, AIMS65, and Rockall score (Fig. 2). A subgroup analysis was performed on the performance of the new risk score for stages I-III patients and stage IV patients who had curative stage but were clinically inoperable. In the case of the stages I-III group (n=37), the prediction of mortality using this was statistically significant (AUC, 0.80; 95% CI, 0.58 to 1.00; p=0.033). In the case of the stage IV group (n=156), the prediction of mortality was also significant (AUC, 0.79; 95% CI, 0.72 to 0.87; p<0.001). In predicting rebleeding, AUC values of the GBS score, Rockall score, and AIMS65 did not show statistical significance. In predicting need for intervention, only Rockall score had a statistically significant 64% predictive power (p=0.001).

Figure 2. Prediction of 30-day mortality. The new score was superior to the other scoring systems.
GB, Glasgow-Blatchford; AUC, area under the curve; CI, confidence interval.

In this study, we developed a new scoring system that could predict mortality in UGI cancer bleeding patients based on risk factors (altered mental status, renal failure, low albumin level, older age, and rebleeding) affecting patient mortality. This new scoring system showed a better prediction for 30-day mortality with an AUC value of 0.79 than existing UGI bleeding risk scoring systems such as GBS, Rockall, and AIMS65 scoring systems. It was important to know when renal failure and altered mental status in our study population have appeared. These were also important risk factor in the non-cancer patients.^13,14,22,23 Most of the cases of renal failure were dialysis patients before bleeding, and most of the changes in mentality appeared after bleeding. Therefore, patients with renal failure or altered mentality after bleeding had a risk for 30-day mortality in our study.

In cancer-related bleeding, our study found that more than half of patients with UGI cancers had chronic or latent bleeding, similar to previous studies.^3,4 There have not been many studies on the predicting power of existing bleeding risk scoring systems in cancer bleeding patients. The GBS has been shown to outperform other pre-endoscopic risk scoring systems when assessing the need for intervention or risk of 30-day death, with AUROC over 0.80.^10,24,25 Therefore, guidelines recommend the use of GBS in bleeding risk assessment in the management of UGI bleeding.^5-8 However, in cancer-related bleeding, prediction power of GBS was found to be as low as AUROC 0.57 for need for intervention and 0.61 of AUROC for 30-day mortality.¹⁷ Therefore, GBS appears to have a low predictive value for cancer-related bleeding, unlike the guidelines. This is probably because GBS does not take malignant information into account when calculating the score.

The Rockall score and AIMS65 are scoring systems developed to predict mortality.^13,14 They have a predictive power of over 80% for mortality in patients with UGI bleeding.^20,26,27 However, in cancer-related bleeding, their predictive power was only about 65% to 70% in this study. Even so, it has a relatively better predictive power than GBS. This is thought to be due to the differences in factors considered when calculating each score. The Rockall score considers comorbidity and disseminated malignant disease in its calculation.¹⁴ However, AIMS65 does not consider malignant lesions. Instead, it takes into account age 65 and albumin level, both of which were confirmed as risk factors for mortality in our study.

Tumor stage can be an important risk factor for mortality. However, our data did not show statistical significance. This seemed to be associated with stronger risk factors associated with patients’ general condition like old age and underlying diseases. According to the epidemiology of gastric cancer in Korea, the overall 5-year relative survival rate of gastric cancer was 77.0%.²⁸ However, the median overall survival in our study was only 3 months. This is due to the high proportion of advanced stage and poor general condition in our study subjects. There are also research results that the more advanced the stage, the more serious gastrointestinal bleeding occurs.²⁹ Appropriate cancer bleeding hemostatic techniques are also constantly being researched. In our study, there was no difference in rebleeding rate according to the endoscopic treatment modality. It is still challenging to find an appropriate method of hemostasis for patients with cancer-related bleeding. In this regard, multidisciplinary approaches of surgical, medical, and radiological oncologists are also being studied.³⁰

This study is distinctive from prior studies in terms of patient inclusion criteria. In a previous study, the patient group included all UGI bleeding patients with an active cancer. Nearly half of patients in that study had a non-cancerous bleeding.³¹ In another retrospective study, only patients with inoperable gastric cancer bleeding were studied.¹⁷ In the present study, among causes with UGI bleeding, those with malignant bleeding were included. Thus, patients with various cancer bleeding were included, such as those with bleeding due to primary gastric cancer, metastatic cancer lesions, esophageal cancer, or duodenal cancer. For pancreatobiliary cancer, patients with direct invasion in the UGI tract and bleeding in this region were chosen. It is true that our study population included a wide variety of cancers, making tumor behavior very heterogeneous. However, we think that the physician should focus that the cause of bleeding is cancer rather than specific cancers in patients with suspected tumor bleeding. Therefore, it is most appropriate for selecting patients related to UGI cancer bleeding.

Our study has the following strengths. First, this study included a large number of patients with UGI cancer bleeding in a large tertiary hospital and evaluated performances of commonly used bleeding risk scoring systems. Second, this study was not limited to specific cancers such as gastric cancer, but all UGI tract tumor bleeding patients. Third, compared to existing bleeding risk scoring systems, the new scoring system showed a higher mortality prediction value.

However, this study has several limitations. First, as this was a single-center study, there might be concerns about generalization. Second, since our cases were obtained retrospectively, selection bias could not be avoided. Third, no validation of the new scoring system was conducted in this study. Therefore, further studies are needed to validate the new scoring system as a predictor of mortality in UGI cancer patients. Fourth, when performing diagnostic EGD, whether to try endoscopic treatment or request non-endoscopic hemostasis immediately depends on the level of endoscopists’ experience.

In conclusion, the new scoring system combined with risk factors identified in this study increased the predictability of mortality in patients with UGI cancer bleeding. These findings suggest that the new scoring system can help guide clinical decision making and improve patient outcomes in this high-risk population. Based on this scoring system, modifiable factors through transfusion, fluid resuscitation, and albumin replacement should be considered for reducing patient mortality. Further validation of the new scoring system and innovative interventions are needed to reduce rebleeding and improve patient outcomes.

This work was funded and supported by the National Research Foundation of Korea (NRF-2019R1A5A2027588 and 2022R1A2C2008281).

Supplementary materials can be accessed at https://doi.org/10.5009/gnl230069.

No potential conflict of interest relevant to this article was reported.

Study concept and design: H.M.K., D.K., J.M.P. Data acquisition: H.M.K. Data analysis and interpretation: H.M.K., D.K. Drafting of the manuscript: H.M.K., J.M.P. Critical revision of the manuscript for important intellectual content: H.M.K., D.K., J.M.P. Statistical analysis: H.M.K., D.K. Obtained funding: J.M.P. Administrative, technical, or material support; study supervision: J.M.P. Approval of final manuscript: all authors.