Alarm from the peritoneum: Gastrointestinal perforations in cancer patients, with a focus on perforations of unknown origin

Abstract

BACKGROUND

Gastrointestinal perforation is a rare but potentially life-threatening complication in cancer patients, often occurring in a complex clinical setting characterized by advanced disease, comorbidities, and ongoing systemic treatments. Despite its clinical relevance, data describing the characteristics, prognosis, and underlying causes of gastrointestinal perforation in oncologic populations remain limited, particularly for events not clearly attributable to tumor involvement or treatment-related factors. We hypothesized that gastrointestinal perforations in cancer patients represent a heterogeneous condition and that cases without an identifiable cause constitute a distinct subgroup with specific clinical characteristics and prognostic implications.

AIM

To characterize cancer patients with gastrointestinal perforation and to estimate overall survival (OS) and associated prognostic factors.

METHODS

This single-center retrospective observational study included adult cancer patients who experienced gastrointestinal perforation between October 2007 and October 2023 at IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST) “Dino Amadori” of Meldola, Italy. Cases were identified through electronic health records and manually validated. Perforations were classified as cancer-related, cancer-unrelated, or of unknown cause based on a review of clinical records, radiological findings, recent procedures, and, when available, surgical and histopathological data. OS and prognostic factors were analyzed using Kaplan-Meier estimates and Cox regression models.

RESULTS

Among 72143 cancer patients, 119 (0.16%) experienced gastrointestinal perforation: 80 (67.2%) cancer-related, 28 (23.6%) cancer-unrelated, and 11 (9.2%) of unknown cause. Colorectal cancer predominated in cancer-related cases (41.3%), while breast (17.9%) and lung cancer (14.3%) in cancer-unrelated perforations; kidney and lung cancers in unknown-cause perforations (27.3% each). The colon (27.5%) and small intestine (23.8%) were the most common perforation sites in cancer-related cases, rectosigmoid (60.7%) in cancer-unrelated cases, and colon (45.5%) in unknown-cause cases. Patients with perforation of unknown cause showed a higher prevalence of comorbidities, particularly arterial hypertension (63.6%). Median OS was 12.1 months, with 30-day mortality of 20.5%. Hypertension and metastatic burden were associated with worse prognosis, while colorectal, breast, and hematologic malignancies showed improved survival.

CONCLUSION

Gastrointestinal perforation is an uncommon but severe event in cancer patients, with a distinct subset of cases occurring without an identifiable cause and with a prognosis influenced by comorbidities and metastatic burden.

Key Words: Gastrointestinal perforation; Spontaneous perforation; Cancer patients; Prognostic factors; Emergency surgery; Overall survival; Real-world evidence

Core Tip: In this single-center retrospective study, a small but clinically relevant subset of gastrointestinal perforations occurred without an identifiable cause. Patients with perforation of unknown origin showed distinctive clinical and pathological features compared with cancer-related and cancer-unrelated cases, including a higher burden of comorbidities, particularly arterial hypertension, a different distribution of primary tumors (lung and kidney cancers), and preferential involvement of the colon and small intestine. Although outcomes appeared broadly comparable across groups, these findings suggest that patient-related factors may contribute to spontaneous perforation, highlighting an under-recognized clinical scenario in oncology practice.

INTRODUCTION

Perforation of the gastrointestinal tract is an abdominal emergency requiring prompt diagnosis and treatment. Unexplained abdominal pain and occasionally back pain when the perforation extends into the retroperitoneum represent a diagnostic challenge for clinicians, as they may conceal a life-threatening condition if not recognized and treated early.

Gastrointestinal perforation can result from numerous conditions, including malignancies. These may cause intestinal obstruction and bowel distension, while lymphoma treated with chemotherapy can lead to bowel wall necrosis. Other causes include diverticulitis, appendicitis, peptic ulcer disease, inflammatory bowel disease, drug-induced ulcers, and invasive procedures. The possibility of perforation should also be considered in patients with chronic constipation or those who have been bedridden for prolonged periods. Awareness of these risk factors is essential for early diagnosis.

Computed tomography scanning can often, though not consistently, localize the site of perforation by identifying extraluminal gas, contrast material leakage, or discontinuity of the gastrointestinal tract[1]. Management of acute abdomen in cancer patients may be particularly complex because of their compromised clinical condition and/or altered hematological status. In some cases, a perforation is identified without a clear underlying cause; these events are classified as spontaneous perforations. Spontaneous gastrointestinal perforation may occur in the stomach or bowel wall of patients receiving systemic chemotherapy and/or corticosteroids, even in the absence of tumor involvement at the perforation site[2]. Therefore, our rationale was to describe the patterns in the etiology of perforations and collect data on their management and on their impact in terms of patient survival. In detail, this study primarily aimed to characterize cancer patients who experienced gastrointestinal perforation in terms of demographic, oncologic, and clinical characteristics, with a particular focus on cases with perforation of unknown etiology.

Furthermore, overall survival (OS) was estimated, and clinical and oncologic factors associated with poorer or improved survival were explored.

MATERIALS AND METHODS

Study design and patients

This study was a single-center, retrospective observational analysis conducted among cancer patients with gastrointestinal perforation at the IRCCS Istituto Romagnolo per lo Studio dei Tumori (IRST) “Dino Amadori” of Meldola, Italy. This investigation was approved by the Comitato Etico della Romagna (Romagna Ethics Committee; Approval No. 1517).

The eligible population included all adult cancer patients who experienced gastrointestinal perforation between October 1, 2007, and October 1, 2023. Potentially eligible patients were identified through a structured search of clinical documents, including clinical notes, medical and radiology reports, and discharge letters. Specifically, the search strategy included terms related to gastrointestinal perforation, such as “free air in the abdomen” and “perforation”, to identify patients with a documented history of abdominal perforation. Oncologists manually reviewed all identified cases to confirm eligibility and classify perforations into three etiologic categories: Cancer-related, cancer-unrelated, or of unknown cause. The date of gastrointestinal perforation was defined as the index date. Patients were followed from the index date until death, the last documented follow-up, or the end of the observational period (August 2024), whichever occurred first. Mortality data were obtained and verified from the Emilia-Romagna region’s mortality register database.

Data collection

The eligible population and patient information were retrospectively retrieved from the electronic health records (EHR; CCE Log80 2.6, Log80 S.r.l.), which are routinely used in clinical practice. The EHR system contains data on clinical visits, routine laboratory tests, disease assessments, administered therapies, and diagnostic and therapeutic procedures performed in both outpatient and inpatient settings.

Demographic, oncologic, and clinical characteristics were collected, including age and sex; primary tumour site; metastatic status and sites; number of metastatic lesions; anatomical site of perforation; surgical management; comorbidities; presence of active infection; recent use of corticosteroids and antibiotic therapy; and chemotherapy exposure.

Data extraction was followed by data cleaning and complemented by a manual review of unstructured data to achieve high data completeness.

Outcomes and definitions

Patients were retrospectively classified into three groups: Cancer-related, cancer-unrelated, or of unknown cause based on the presumed etiology of gastrointestinal perforation. For each patient, clinical records were systematically reviewed by a physician, integrating medical history, imaging findings, treatment exposure, procedural history, and, when available, surgical and histopathological reports. Predefined etiological variables, including tumor site perforation, post-chemotherapy perforation, intestinal obstruction, diverticulitis, iatrogenic perforations, and gastric ulcer, were assessed and coded as dummy variables (0 = absent, 1 = present).

Cancer-related perforations included events attributable to intestinal obstruction caused by tumour burden, tumour site perforation, and post-chemotherapy perforation. Specifically, perforations were considered chemotherapy-related when occurring within 7 days of the last administration of systemic treatment, particularly in patients receiving drugs known to be associated with increased perforation risk (e.g., antiangiogenic agents). Evidence of bowel obstruction, perforation adjacent to the tumour, peritoneal carcinomatosis, or histological confirmation of neoplastic infiltration supported classification as cancer-related.

Cancer-unrelated perforations comprised perforations associated with diverticulitis, gastric ulcer disease and iatrogenic perforations. Diverticulitis- and peptic ulcer-related perforations were assigned based on imaging findings and/or clinical history. Iatrogenic perforations were identified when they occurred within 7 days of recent endoscopic or interventional procedures (biliary drainage and enema).

Perforations were classified as of unknown cause when no predefined etiological factors were identified after the structured review and all assessed variables were coded as absent.

OS was defined as the time from the date of perforation (i.e., index date) to death from any cause or last follow-up, whichever occurred first.

Time-to-surgery was defined as the time interval between gastrointestinal perforation and surgical intervention. In practice, patients without a recorded surgery date were censored at the date of death or the last follow-up, whichever occurred first.

The time from chemotherapy to gastrointestinal perforation was calculated only in patients who received chemotherapy and was defined as the interval between the last chemotherapy administration and the perforation.

Statistical analysis

Descriptive statistics were reported for each collected variable. More precisely, continuous variables were summarized using mean ± SD or median [interquartile range (IQR)], as appropriate, while categorical variables were reported as absolute n (%). Patient characteristics were described according to perforation etiology (cancer-related, cancer-unrelated, or unknown cause). No formal hypothesis testing on group differences in patients’ characteristics was planned, given the descriptive and exploratory nature of the study.

Survival distributions were estimated using the Kaplan-Meier method and compared across perforation groups using the log-rank test. Median OS and survival probabilities at predefined time points (30-day survival) were reported with corresponding 95% confidence intervals (95%CIs).

Furthermore, the association between clinical and treatment-related variables and OS was explored using the Cox proportional hazards regression model. Practically, a set of demographic and clinical characteristics was initially evaluated as potential prognostic factors, including comorbidities, tumour characteristics, perforation site and metastatic localizations. To avoid model overfitting given the limited number of events, variable selection was performed using a backward stepwise procedure based on the Akaike Information Criterion (AIC). Coherently, the final Cox proportional hazards model included only variables retained after this selection process. Hazard ratios with 95%CIs were estimated and displayed using forest plots.

Moreover, time-to-event analyses were also used to evaluate: (1) The interval between gastrointestinal perforation and surgical intervention; and (2) The timespan between the last chemotherapy administration and the diagnosis of gastrointestinal perforation, when applicable. These time-to-event variables were analyzed using Kaplan-Meier methods and stratified by perforation etiology.

Missing data were considered to be completely at random and, consequently, handled using an available-case approach. No multiple imputation procedures were applied given the descriptive nature of the study and the relatively limited proportion of missing values on key variables. Accordingly, variables with substantial missingness were used only for descriptive purposes and were not included in the multivariable Cox regression model.

Statistical analysis were performed by a biomedical statistician using R statistical software (www.r-project.org) version 4.4.2.

RESULTS

Patient’s characteristics at baseline

Among 72143 seen or treated at our cancer centre during the study period, 119 (0.16%) experienced a gastrointestinal perforation and were analyzed. Of these, 108 (90.8%) cases had an identifiable cause of perforation, while 11 (9.2%) cases were classified as having an unknown cause (Figure 1). Patients with a known cause of perforation were further categorized into cancer-related (n = 80, 74.1%) and cancer-unrelated groups (n = 28; 25.9%).

Figure 1 Patients’ disposition. Classifications of gastrointestinal perforation in the study population according to etiology, including cancer-related, cancer-unrelated and unknown causes. ¹Iatrogenic perforations include those occurring within 7 days after invasive procedures such as gastroscopy, colonoscopy, biliary drainage, or enema.

The average follow-up after gastrointestinal perforation was 24.0 months, whereas the median was 7.2 (IQR: 1.5-39.0) months.

The demographic and clinical characteristics of the patients are shown in Table 1. The mean age at cancer diagnosis was slightly lower in the cancer-related perforation group than in the other two groups. Male patients were slightly more prevalent in all perforation categories. In the cancer-related perforation group, the most frequent primary malignancy was colorectal cancer (41.3%). In contrast, breast cancer (17.9%) and lung cancer (14.3%) were among the most common primary tumors in the cancer-unrelated perforation group. In patients with perforation of unknown cause, kidney and lung cancers (27.3% each) were the most frequently observed primary malignancies.

Table 1 Patients’ characteristics according to perforation etiology.

Characteristics	Cancer-related cause (n = 80)	Cancer-unrelated cause (n = 28)	Unknown cause (n = 11)
Age at diagnosis (year)	63.0 ± 12.22	66.7 ± 9.73	65.9 ± 8.99
Sex
Female	38 (47.5)	12 (42.9)	4 (36.4)
Male	42 (52.5)	16 (57.1)	7 (63.6)
Primary tumor site
Colon	33 (41.3)	3 (10.7)	1 (9.1)
Hematological cancer	3 (3.8)	3 (10.7)	1 (9.1)
Breast	0 (0.0)	5 (17.9)	1 (9.1)
Pancreas	7 (8.8)	3 (10.7)	0 (0.0)
Lung	0 (0.0)	4 (14.3)	3 (27.3)
Kidney	0 (0.0)	2 (7.1)	3 (27.3)
Rectum	9 (11.3)	1 (3.6)	1 (9.1)
Multiple sites	1 (1.3)	0 (0.0)	0 (0.0)
Stomach	13 (16.3)	2 (7.1)	0 (0.0)
Other sites	14 (17.5)	5 (17.9)	1 (9.1)
Metastatic disease
No	24 (30.0)	9 (32.1)	4 (36.4)
Yes	55 (69.8)	19 (67.9)	7 (63.6)
Unknown	1 (1.3)	0 (0.0)	0 (0.0)
Number of metastatic sites
1	33 (60.0)	13 (68.4)	3 (42.9)
2	17 (30.9)	2 (10.5)	3 (42.9)
3	5 (9.1)	3 (15.8)	0 (0.0)
4	0 (0.0)	1 (5.3)	1 (14.3)
Site of metastases
Lung	13 (16.3)	3 (10.7)	3 (27.3)
Lymph nodes	10 (12.5)	7 (25.0)	3 (27.3)
Bone	2 (2.5)	9 (32.1)	3 (27.3)
Peritoneum	27 (33.8)	2 (7.1)	0 (0.0)
Brain	3 (3.8)	1 (3.6)	1 (9.1)
Adrenal gland	0 (0.0)	0 (0.0)	1 (9.1)
Skin	1 (1.3)	0 (0.0)	0 (0.0)
Stomach	2 (2.5)	1 (3.6)	0 (0.0)
Liver	19 (23.8)	3 (10.7)	2 (18.2)
Ovary	1 (1.3)	0 (0.0)	0 (0.0)
Uterus	1 (1.3)	0 (0.0)	0 (0.0)
Pancreas	1 (1.3)	0 (0.0)	0 (0.0)
Kidney	1 (1.3)	1 (3.6)	0 (0.0)
Colon	0 (0.0)	1 (3.6)	0 (0.0)
Muscle	1 (1.3)	1 (3.6)	0 (0.0)
Site of perforation
Gallbladder	0 (0.0)	1 (3.6)	0 (0.0)
Colon	22 (27.5)	2 (7.1)	5 (45.5)
Small intestine	19 (23.8)	3 (10.7)	3 (27.3)
Rectosigmoid	16 (20.0)	17 (60.7)	1 (9.1)
Unknown	14 (17.5)	4 (14.3)	2 (18.2)
Stomach/cardia	8 (10.0)	1 (3.6)	0 (0.0)
Bladder	1 (1.3)	0 (0.0)	0 (0.0)
Surgery
No	30 (37.5)	9 (32.1)	3 (27.3)
Yes	50 (62.5)	19 (67.9)	7 (63.6)
Unknown	0 (0.0)	0 (0.0)	1 (9.1)
Surgical technique
Resection + stoma	21 (42.0)	7 (36.8)	3 (42.9)
Stoma only	2 (4.0)	1 (5.3)	0 (0.0)
Resection	25 (50.0)	8 (42.1)	4 (57.1)
Drainage only	1 (2.0)	1 (5.3)	0 (0.0)
Others	1 (2.0)	2 (10.5)	0 (0.0)
Time to surgery (day)	1.0 (1-10)	1.5 (1-22)	1.0 (10-23)
Comorbidities
No comorbidities	44 (55.0)	10 (35.7)	2 (18.2)
Inflammatory bowel disease	1 (1.3)	1 (3.6)	0 (0.0)
Arterial hypertension	20 (25.0)	9 (32.1)	7 (63.6)
Ischemic heart disease	1 (1.3)	0 (0.0)	2 (18.2)
Renal insufficiency	2 (2.5)	1 (3.6)	0 (0.0)
Rheumatologic disease	1 (1.3)	1 (3.6)	0 (0.0)
Metabolic syndrome	7 (8.8)	4 (14.3)	1 (9.1)
Aortic aneurysm	2 (2.5)	0 (0.0)	0 (0.0)
COPD	4 (5.0)	1 (3.6)	0 (0.0)
Atrial fibrillation	3 (3.8)	1 (3.6)	0 (0.0)
Hypothyroidism	5 (6.3)	2 (7.1)	1 (9.1)
Active infection
No	69 (86.3)	24 (85.7)	9 (81.8)
Yes	1 (1.3)	1 (3.6)	0 (0.0)
Unknown	10 (12.5)	3 (10.7)	2 (18.2)
Corticosteroid therapy within 7 days before perforation
No	18 (22.5)	7 (25.0)	3 (27.3)
Yes	22 (27.5)	13 (46.4)	5 (45.5)
Unknown	40 (50.0)	8 (28.6)	3 (27.3)
Chemotherapy
With known perforation risk	20 (25.0)	9 (32.1)	4 (36.4)
Without known perforation risk	26 (32.5)	9 (32.1)	5 (45.5)
No chemotherapy	34 (42.5)	10 (35.7)	2 (18.2)
Antibiotic therapy within 7 days before perforation
No	33 (41.3)	7 (25.0)	5 (45.5)
Yes	2 (2.5)	7 (25.0)	0 (0.0)
Unknown	45 (56.3)	14 (50.0)	6 (54.5)

Continuous variables were summarized using means ± SD or medians (interquartile ranges), as appropriate, while categorical variables were reported as absolute n (%). COPD: Chronic obstructive pulmonary disease.

The proportion of patients with metastatic disease was similar across the three groups, with approximately two-thirds presenting with metastases. In the cancer-related perforation group, the peritoneum was the most common site of metastasis (33.8%), whereas bone metastases were most prevalent in the cancer-unrelated group (32.1%). Among patients with perforation of unknown origin, metastatic sites were mainly distributed among the lungs, lymph nodes, and bones (each 27.3%).

Regarding the anatomical site of perforation, the colon (27.5%) and small intestine (23.8%) were the most frequently involved sites in the cancer-related group. In comparison, rectosigmoid perforation was predominant in the cancer-unrelated group (60.7%). Among patients with perforation of unknown cause, the colon was the most common site of perforation (45.5%).

Overall, the majority of patients underwent surgical management following gastrointestinal perforation, with similar rates across the three groups (62.5% in the cancer-related group, 67.9% in the cancer-unrelated group, and 63.6% in the unknown-cause group). Surgical bowel resection with or without stoma creation represented the most common operative approach in all three groups. Notably, only 18.2% of patients with perforation of unknown cause had no documented comorbidities, compared with 55.0% of patients in the cancer-related perforation group.

Corticosteroid therapy within the seven days preceding perforation was relatively common, particularly in the cancer-unrelated (46.4%) and unknown-cause (45.5%) groups. Chemotherapy administration, including regimens not traditionally associated with a high risk of perforation, was frequently observed across all groups (32.5% in the cancer-related group, 32.1% in the cancer-unrelated group, and 45.5% in the unknown-cause group). Antibiotic therapy within 7 days of perforation was also frequently reported, although data were missing for a substantial proportion of patients in all groups.

OS

OS after gastrointestinal perforation was similar among the three groups (Figure 2). All groups showed a marked early mortality following perforation, with survival curves over time largely overlapping. More precisely, the proportion of survivors 30 days after cancer-related perforation was 80.9% (95%CI: 72.6%-90.1%). Similarly, in the cancer-unrelated perforation group, the estimated 30-day survival was 78.6% (95%CI: 64.8%-95.3%). Notably, in the unknown origin perforation group, the proportion of 30-day survivors was estimated at 72.7% (95%CI: 50.6%-100.0%). The median OS was comparable among the three groups (Figure 2), and no statistically significant differences were observed (log-rank test P = 0.954). Specifically, among patients who experience cancer-related perforation, the median OS was 12.1 (95%CI: 3.5-28.4) months. Similarly, in the cancer-unrelated perforation group, the estimated median OS was 11.2 months (95%CI: 3.8-38.1), whereas in the unknown origin of perforation group, the median OS was 17.2 (95%CI: 2.6-not reached) months.

Figure 2 Overall survival. Kaplan-Meier curves of overall survival from the time of gastrointestinal perforation, stratified by perforation etiology (cancer-related, cancer-unrelated, and unknown cause). Log-rank test P = 0.954. Numbers at risk are shown below. OS: Overall survival.

Baseline factors associated with survival

Prognostic factors for survival were explored using a forest plot analysis (Figure 3). Arterial hypertension (P = 0.012), bone metastases (P < 0.001), peritoneal carcinomatosis (P < 0.001), and liver metastases (P < 0.001) were significantly associated with worse prognosis. Conversely, patients with colorectal (P < 0.001), breast (P < 0.001), or hematological malignancies (P = 0.025) showed improved survival (Figure 3).

Figure 3 Forest plot. Forest plot showing hazard ratios and 95% confidence intervals for overall survival after gastrointestinal perforation. Analyses were performed using Cox proportional hazards models. The reference categories are indicated. A hazard ratio greater than 1 indicates worse prognosis, while a hazard ratio less than 1 indicates improved survival. P values are reported on the right.

The time interval between the last chemotherapy and the diagnosis of gastrointestinal perforation was short and comparable across all three groups (Figure 4A). Regarding surgical timing, most patients underwent surgery within the first or second day after diagnosis (Figure 4B).

Figure 4 Time interval. A: From the last chemotherapy to gastrointestinal perforation. Kaplan-Meier curves showing the time interval from the last chemotherapy administration to gastrointestinal perforation, stratified by perforation etiology (cancer-related, cancer-unrelated, and unknown cause). The analysis includes only patients who received chemotherapy prior to perforation; B: From gastrointestinal perforation to surgical interventions. Kaplan-Meier curves showing the time from gastrointestinal perforation to surgical intervention, stratified by perforation etiology (cancer-related, cancer-unrelated, and unknown cause). The analysis includes only patients who underwent surgery. Numbers at risk are shown below the X-axis.

DISCUSSION

We present a single-center, retrospective analysis of gastrointestinal perforations in onco-hematological patients occurring during their cancer care pathways. Three main groups of gastrointestinal perforations were identified: Cancer-related, cancer-unrelated and perforations of unknown origin. This study characterizes cancer patients with gastrointestinal perforation and estimates OS and prognostic factors associated with perforation, demonstrating that cancer patients are frequently considered for emergency surgical management.

Importantly, our findings confirm that gastrointestinal perforation may occur unexpectedly and without a clear underlying cause in clinical oncology practice.

Following perforation, patients typically present with acute abdominal pain; however, if diagnosis and treatment are delayed, the clinical course may rapidly progress to septic shock and multiorgan failure[3]. The clinical significance of peritonitis was already recognized by Hippocrates, who described the so-called Hippocratic facies, which remains a relevant predictive indicator of severe intra-abdominal pathology.

Several older and more recent retrospective studies have investigated gastrointestinal perforation in cancer patients, primarily focusing on clinical characteristics, outcomes, and mortality risk[4-6]. However, these studies are not directly comparable due to differences in study design, cancer patient populations, and perforation sites. In the present study, the incidence of gastrointestinal perforation was estimated at approximately 0.16% in the entire oncologic patient cohort. In contrast, several previously did not report a denominator[4,7], while Hapani et al[8] reported an incidence of 0.9% in bevacizumab-treated patients, using a drug-specific trial population as the denominator. Similarly, Mazepa et al[5] reported a very low incidence of approximately 0.03% among patients receiving chemotherapy, based on a treatment-session denominator. Previous studies also adopted different inclusion criteria and definitions of perforation. Torosian and Turnbull[4] excluded instrumental perforations, anastomotic leaks, primary perforating adenocarcinoma, and obstruction-related perforations, defining spontaneous perforation as events occurring during chemotherapy and/or corticosteroid therapy without specifying the interval between treatment and perforation. Mazepa et al[5] included only patients with intraoperative gastroduodenal perforation and defined active treatment as anticancer therapy administered within 60 days before perforation. Maeda et al[6] analyzed patients undergoing emergency surgery for acute abdomen within 30 days after anticancer drugs administration. In the study by Mazepa et al[5], more than half of the patients had metastatic disease, whereas the extent of disease was not reported in the studies by Maeda et al[6] and Torosian and Turnbull[4]. Conversely, our study included all gastrointestinal perforations in an unselected oncologic population and classified them by presumed etiology, showing a predominance of cancer-related events and identifying a subgroup without a clear cause. The anatomical distribution of perforations differed across studies. Mazepa et al[5] included only gastroduodenal perforations, whereas Torosian and Turnbull[4] reported a predominance of small intestine (45%) and colonic perforations (55%). Maeda et al[6] reported gastrointestinal perforations involving different segments of the gastrointestinal tract, including the stomach, small intestine, colon, and appendix, indicating a heterogeneous anatomical distribution. In contrast, gastric perforations represented only a small proportion of cases (10%) in our cohort and were mainly observed in the cancer-related group. Rectosigmoid perforation predominated in the non-cancer-related group (60.7%), largely associated with diverticulitis. These differences highlight the heterogeneity of perforation patterns across oncologic populations and reflect the different clinical settings and inclusion criteria of previous studies. Outcomes also varied substantially across studies. Mazepa et al[5] reported a high mortality rate among oncologically treated patients with gastroduodenal perforation, with a 30-day mortality of 55.6%, regardless of recent chemotherapy exposure. In Maeda et al[6], 30-day mortality was 26%, whereas the operative mortality reported by Torosian and Turnbull[4] was 53%. In our cohort, the 30-day mortality was 20.5%, slightly lower than that reported by Maeda et al[6]. Differences in mortality across these studies are likely due to differences in patient selection and clinical settings. Interestingly, the estimated proportion of 30-day survivors was 80.1%, 78.6% and 72.7% for cancer-related perforations, cancer-unrelated perforations and the unknown origin of the perforation group, respectively. Mortality and morbidity associated with spontaneous perforation are likely exacerbated by malnutrition, pancytopenia, and immunosuppression commonly observed in patients undergoing chemotherapy.

In terms of prognostic factors, Maeda et al[6] identified older age, poor performance status, and low serum albumin as predictors of in-hospital death after emergency surgery during chemotherapy. In contrast, Mazepa et al[5] found age ≥ 65 years to be the only independent predictor of 30-day mortality, and Torosian and Turnbull[4] did not identify clear prognostic factors. Our analysis identified arterial hypertension and the presence of bone, peritoneal, or hepatic metastases as negative prognostic factors. Although the underlying mechanisms remain unclear, hypertension may contribute to microvascular damage and impaired intestinal perfusion, potentially weakening bowel wall integrity and increasing susceptibility to perforation.

The role of chemotherapy as a contributing factor to gastrointestinal perforation remains controversial. In the study by Mazepa et al[5], only 35.5% of patients received chemotherapy, whereas all patients in the cohort reported by Maeda et al[6] were undergoing chemotherapy at the time of perforation. Torosian and Turnbull[4] reported that 76.6% of patients received corticosteroids alone or in combination with chemotherapy. In our cohort, fewer than half of patients were not receiving chemotherapy at the time of perforation, with relevant differences among the three groups (42.5%, 35.7%, and 18.2% in the cancer-related, cancer-unrelated, and unknown-cause groups, respectively). Notably, the time interval between the last chemotherapy administration and the perforation diagnosis was short and comparable across all groups, suggesting a temporal association between systemic treatment and perforation onset. The potential role of concomitant medications has also been explored[9,10]. Non-steroidal anti-inflammatory drugs and corticosteroids have both been implicated as risk factors for gastrointestinal perforation. The association with corticosteroid therapy was first reported by Beck et al[11] in 1950. Although the exact mechanism remains unclear, steroids are thought to impair the intestinal mucosal barrier, possibly through inhibition of prostacyclin synthesis. Although corticosteroid use was more frequent (37.8%), no correlation with perforations was observed[5]. In our study, corticosteroid use in the week preceding perforation was frequent across all groups, although less prevalent than expected in the cancer-related group.

The literature indicates an increased risk of gastrointestinal perforation in patients treated with biologic and antiangiogenic agents, particularly bevacizumab, aflibercept, ramucirumab and regorafenib[8,12-14]. In contrast, the association between “classic” cytotoxic chemotherapy and perforation is less clear. While case reports describe perforation events, a consistent increase above baseline risk has not been uniformly demonstrated. Potential mechanisms include tumor infiltration, rapid tumor necrosis, severe mucositis, pre-existing bowel disease, and synergistic effects with radiotherapy, especially in the upper gastrointestinal tract[15,16]. In intestinal lymphomas, rapid tumor lysis following chemotherapy is frequently implicated as a primary mechanism[17]. Furthermore, severe neutropenia and neutropenic enterocolitis represent particularly high-risk conditions, as mucosal injury and ischemic necrosis may lead to single or multiple perforations, especially in pediatric patients or those with leukemia[18].

While perforations related to chemotherapy or targeted agents may be clinically explicable in cancer-related cases, perforations of unknown origin remain more difficult to interpret. Early reports from the 1970s already described spontaneous gastrointestinal perforation in oncology patients. Lundy et al[7] reported 36 cases, of which five perforations occurred in the absence of tumor involvement at the perforation site. In such cases, perforation is considered spontaneous.

Subsequent studies, including the experience from Memorial Sloan-Kettering Cancer Center described the absence of tumour infiltration in nearly half of surgical specimens[4]. More recently, spontaneous intestinal perforation has been described as low in patients with malignant lymphoma treated with chemotherapy, although its true incidence remains unknown[2,19]. In our cohort, 11 patients were classified as having perforation of unknown origin. We preferred to use the term perforation of unknown origin rather than spontaneous perforation, because only surgery or autopsy could definitively rule out the presence of cancer at the perforation site. The absence of cancer was confirmed by histological examination in only 7 of the 11 patients; therefore, these cases could be considered true spontaneous perforations. In the remaining four cases, potential causes of perforation were excluded based on the available clinical information. For this reason, we considered the term unknown origin more appropriate than spontaneous perforation.

Despite the inability to clearly identify the etiopathogenetic mechanisms of perforation, particularly in patients with perforation of unknown origin, our data suggest that patient-related factors may play a relevant role. Indeed, only two patients (18.2%) in the unknown-cause group had no comorbidities, and arterial hypertension was the most frequent comorbidity (63.3%).

Gastrointestinal perforations during active oncologic treatment represent a particularly challenging clinical scenario, especially in elderly patients, those with advanced disease, or those with peritoneal carcinomatosis. Despite these concerns, most patients in our cohort underwent surgical intervention, irrespective of perforation etiology. In line with previous reports, including that of Torosian and Turnbull[4], early surgical management appears to be the most appropriate strategy when feasible.

This study has several limitations. First, its retrospective design may introduce selection and information bias. Second, the study population was heterogeneous with respect to tumor type and treatment regimens. Third, the subgroup of patients with perforation of unknown cause was relatively small (n = 11), which limits statistical power and precludes definitive conclusions about this specific cohort. Fourth, not all cases were confirmed by surgical or histopathological findings, and detailed data on postoperative complications in patients who underwent surgery were not available. Finally, the etiopathogenetic mechanisms underlying perforations of an unknown origin could not be determined. Furthermore, from a methodological point of view, although several demographic and clinical variables were explored during the model-developing phase, the final multivariable Cox regression included only variables retained after a data-driven selection procedure (i.e., based on AIC). Important prognostic factors in oncology (e.g., performance status, nutritional status) were not consistently available in the EHR and therefore could not be incorporated into the analysis. Consequently, the observed associations should be interpreted cautiously, as residual confounding from unmeasured clinical factors cannot be excluded.

Despite these limitations, our study has notable strengths. It represents a large analysis of gastrointestinal perforation in a broad oncologic population. Importantly, the inclusion of an unselected cohort of cancer patients identified from the entire population accessing our institution over a long observation period reflects real-world clinical practice and may enhance the generalisability of our findings to similar oncology settings. Additionally, it expands current knowledge on prognostic factors beyond those previously reported[5,6]. These findings may help identify patients most suitable for surgical management and constitute one of the largest single-center experiences of spontaneous perforation after the study by Torosian and Turnbull[4] and Lundy et al[7]. Although the incidence of perforation is low (and even lower for those of unknown origin), we believe it is important to remain vigilant even in apparently “stable” patients (e.g., those with arterial hypertension, bone metastases, or liver metastases). Furthermore, maintaining regular bowel movements may represent good clinical practice.

CONCLUSION

Arterial hypertension, bone metastases, peritoneal carcinomatosis and liver metastases are associated with worse prognosis in patients with gastrointestinal perforations. Perforation often occurred shortly after chemotherapy administration and frequently warrants surgical considerations, even in patients with advanced malignancy. Clinicians should be aware that unexpected and apparently unjustified gastrointestinal perforation may occur in oncology practice.