Peritoneal drainage versus primary laparotomy in premature newborns with intestinal perforation: a retrospective cohort study evaluating an etiology-driven surgical approach

Abstract

Background

The surgical management of intestinal perforation in extremely low birth weight (ELBW) infants, stemming from necrotizing enterocolitis (NEC) or spontaneous intestinal perforation (SIP), is a subject of enduring debate. The comparative effectiveness of initial bedside peritoneal drainage (PD) versus formal primary laparotomy (PL) on mortality and major morbidity is not fully resolved.

Objective

To compare 90-day mortality and comprehensive morbidity outcomes between PD-first and PL-first strategies in a homogeneous cohort of preterm infants with intestinal perforation, with rigorous subgroup analysis based on disease etiology (NEC vs. SIP).

Methods

A single-center retrospective cohort study (2010–2023) included infants with gestational age < 32 weeks and/or birth weight < 1500 g undergoing surgery for pneumoperitoneum. Propensity score matching (PSM) with comprehensive illness severity covariates was used to balance baseline characteristics. Primary outcome was 90-day all-cause mortality, with secondary analysis of attributable versus non-attributable deaths. Secondary outcomes included short-term surgical morbidity, nutritional independence, long-term gastrointestinal outcomes, and incidence of short-bowel syndrome (SBS). Sensitivity analyses addressed temporal bias and confounding by non-GI comorbidities.

Results

Of 126 eligible infants, 66 underwent PD-first and 60 PL-first. After PSM with 15 covariates (n = 94), overall mortality was comparable (PD: 34.0% vs. PL: 36.2%, p = 0.81; p-for-interaction between treatment and etiology = 0.02). In the SIP cohort (n = 41), PD was associated with significantly lower mortality (15.8% vs. 45.5%, p = 0.03; adjusted HR 0.32, 95% CI 0.11–0.91) and fewer major surgical complications (21.1% vs. 54.5%, p = 0.03). Among 14 SIP patients managed definitively with PD alone, median 24-month follow-up revealed no cases of intestinal obstruction or stricture. In the NEC cohort (n = 53), PL was associated with a non-significant trend towards lower mortality (38.5% vs. 55.6%, p = 0.24) but a significantly higher rate of surgical complications (57.7% vs. 29.6%, p = 0.04). Time to full enteral feeds was shorter in the PD group overall (median 42 vs. 55 days, p = 0.02), a finding that persisted after excluding patients with non-GI comorbidities. Sensitivity analyses restricted to 2015–2023 confirmed main findings.

Conclusion

The observed association between initial surgical strategy and survival is contingent upon underlying etiology. PD appears advantageous for SIP, was associated with lower mortality and morbidity with acceptable long-term outcomes. For NEC, PL may offer a survival trend advantage despite higher immediate surgical morbidity, likely due to definitive source control. These findings strongly advocate for an etiology-driven surgical decision-making paradigm.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12893-026-03879-z.

Introduction

Intestinal perforation represents a catastrophic abdominal emergency in the premature infant population, with mortality historically exceeding 50% [1]. The two predominant etiologies are necrotizing enterocolitis (NEC), a multifactorial inflammatory necrosis of the intestine, and spontaneous intestinal perforation (SIP), a focal ileal perforation often linked to perinatal stress, glucocorticoid exposure, and early indomethacin use [2]. Despite overlapping clinical and radiographic features—both typically present with pneumoperitoneum and acute deterioration—their underlying pathophysiologies differ substantially, creating a critical junction in management. Distinguishing features (e.g., pneumatosis intestinalis or portal venous gas on imaging suggesting NEC, versus isolated pneumoperitoneum in the absence of these findings suggesting SIP) are often subtle, and definitive differentiation may only be possible intraoperatively.

The surgical debate centers on two approaches: bedside percutaneous peritoneal drainage (PD), a minimally invasive procedure to evacuate enteric contents and control sepsis, and formal exploratory laparotomy (PL) with bowel resection and stoma formation or primary anastomosis. Proponents of PD argue it is better tolerated by physiologically fragile infants, serving as a bridge to later laparotomy or, in some cases, as potentially definitive management in selected cases [3]. Advocates for PL contend that only direct inspection allows for resection of non-viable bowel, adequate source control, and accurate diagnosis [4].

Randomized controlled trials, notably the landmark NECSTEPS trial by Moss et al. [5], found no significant difference in 90-day mortality between PD and PL. However, subsequent meta-analyses have revealed heterogeneity, suggesting that subgroup effects, particularly based on disease etiology, may be obscured [6]. Furthermore, questions remain regarding long-term outcomes of PD as potentially definitive management in selected cases, causes of mortality, and potential confounding by illness severity and temporal changes in practice.

This study aims to provide a detailed, etiology-stratified comparison of PD versus PL in a contemporary, homogeneous cohort of preterm infants, evaluating not only survival but also granular surgical and nutritional morbidity, long-term gastrointestinal outcomes, and causes of death, while rigorously addressing methodological challenges through comprehensive propensity score adjustment, sensitivity analyses, and formal interaction testing.

Methods

Study design and setting

This was an Institutional Review Board-approved retrospective cohort study conducted at a level IV neonatal intensive care unit (NICU) with an annual admission rate of > 80 infants < 28 weeks gestation. The study adhered to the STROBE guidelines for observational studies.

Study population

Infants were identified from a prospectively maintained surgical database. Inclusion criteria were: (1) Gestational age < 32 weeks and/or birth weight < 1500 g, (2) Radiographically confirmed pneumoperitoneum on two-view abdominal X-ray (supine and left lateral decubitus or cross-table lateral view), (3) Surgical intervention (PD or PL) between January 2010 and December 2023. Exclusion criteria were: (1) Congenital gastrointestinal anomalies (e.g., intestinal atresia, meconium ileus), (2) Isolated gastric perforation, (3) Perforation secondary to iatrogenic injury (e.g., rectal thermometer), (4) Treatment limitation/comfort care order in place at time of diagnosis. The inclusion threshold of birth weight < 1500 g (rather than the more restrictive < 1000 g commonly cited in the literature) was selected based on our institutional protocol, which considered PD as an option for all VLBW infants due to comparable physiological fragility, preoperative instability, and high risk of general anesthesia. Additionally, the transition to PD as first-line therapy for presumed SIP after 2015 was applied regardless of weight within the VLBW category.

Exposure and group allocation

The exposure variable was the initial surgical procedure. The PD-first group comprised infants who underwent percutaneous abdominal drain placement (typically 10-French pigtail catheter) at the bedside under local anesthesia as the primary intervention. The PL-first group comprised infants who were transferred to the operating room for exploratory laparotomy via transverse supraumbilical incision as the primary intervention. The choice of procedure was based on surgeon preference and evolving institutional protocol, which increasingly favored PD for presumed SIP after 2015.

Data collection and variable definitions

Demographics & preoperative status

Data extracted from electronic medical records included birth weight, gestational age, sex, postnatal age at diagnosis, antenatal steroid exposure, indomethacin exposure, pre-operative inotropic support (dopamine > 10 mcg/kg/min or epinephrine at any dose), mean arterial pressure, base deficit, preoperative pH, lactate, platelet count, C-reactive protein, and mechanical ventilation parameters (mean airway pressure, FiO₂, oxygenation index) at the time of perforation diagnosis.

Etiology classification

Infants were classified as having NEC if operative findings described pan-intestinal ischemia, necrosis, or if pre-operative imaging demonstrated pneumatosis intestinalis and/or portal venous gas (Bell’s Stage III). Infants were classified as having SIP if operative or PD effluent findings described a focal perforation (typically ileal) without macroscopic necrotic bowel, and pre-operative imaging showed pneumoperitoneum in the absence of pneumatosis [7]. All classifications were adjudicated by two blinded senior neonatologists.

Primary outcome

All-cause mortality within 90 days of the initial surgical procedure. Additionally, we performed a secondary analysis classifying deaths as: (a) attributable to intestinal pathology/surgical management (sepsis from peritonitis, progressive NEC, postoperative complications), or (b) non-attributable (respiratory failure, intraventricular hemorrhage, extreme prematurity, unrelated causes). Classification was performed independently by two senior neonatologists and a pediatric surgeon, with consensus reached for all cases.

Secondary outcomes

Short-term surgical morbidity

Major surgical morbidity was defined as any of the following occurring within 30 days postoperatively, or until hospital discharge if discharge occurred before 30 days (n = 12 patients) : (1) wound infection requiring antibiotic therapy or surgical drainage, (2) wound dehiscence (superficial or fascial), (3) stoma complications including prolapse (> 3 cm), stricture requiring revision, necrosis, or retraction, (4) intra-abdominal abscess confirmed by imaging and requiring drainage, and (5) anastomotic leak confirmed by imaging or at reoperation.

Nutritional outcomes

Time to achieve full enteral feeds was defined as the first day of achieving ≥ 120 mL/kg/day of enteral nutrition for at least 72 consecutive hours without parenteral nutrition supplementation. Duration of parenteral nutrition (PN) was recorded as total days of PN support. Short-Bowel Syndrome (SBS) was defined as PN dependence > 90 days post-operatively with anatomic resection of > 50% of small intestine or clinical evidence of intestinal failure [8].

Procedural outcomes

Need for a second surgical procedure, classified as: (a) unplanned reoperations (rescue laparotomy after PD, reoperation for complications including anastomotic leak, bowel obstruction, or stoma necrosis requiring revision), and (b) planned reoperations (elective stoma closure, scheduled second-look laparotomy). Total number of operations during initial hospitalization was also recorded. PD failure was defined as the need for rescue laparotomy following initial PD. The decision to proceed with rescue laparotomy was based on a composite of clinical triggers occurring within 72 h of PD placement (median time to rescue laparotomy was 48 h, IQR 24–72 h), including: (a) persistent or worsening hemodynamic instability despite maximal medical support (persistent hypotension requiring escalating inotropes), (b) rising arterial lactate (> 2 mmol/L increase from post-PD nadir), (c) worsening metabolic acidosis (pH < 7.20 with base deficit >-10), (d) clinical deterioration with increasing abdominal distension or discoloration, or (e) persistent positive PD effluent cultures after 48 h despite appropriate antibiotic therapy.

Resource utilization

Total length of hospital stay (from birth to discharge or death), and duration of postoperative mechanical ventilation.

Long-term follow-up

For SIP patients managed definitively with PD alone, long-term gastrointestinal outcomes were assessed including intestinal obstruction, stricture, dysmotility, and growth parameters at 12 and 24 months post-discharge.

Statistical analysis

Categorical variables are presented as counts (percentages) and compared using Chi-square or Fisher’s exact test. Continuous variables, assessed for normality using Shapiro-Wilk test, are presented as mean ± SD or median (IQR) and compared using Student’s t-test or Mann-Whitney U test.

Propensity score matching

To address potential selection bias, a propensity score (PS) was developed using multivariable logistic regression with initial procedure (PD vs. PL) as the dependent variable. Covariates included: birth weight, gestational age, sex, postnatal age at diagnosis, preoperative inotrope use, mean arterial pressure, base deficit, mean airway pressure, FiO₂ requirement, oxygenation index, arterial lactate, pH, platelet count, C-reactive protein, antenatal steroid exposure, indomethacin exposure, and etiology (NEC/SIP). A 1:1 nearest-neighbor matching algorithm with a caliper width of 0.2 SD of the logit of the PS was used to create a matched cohort. Covariate balance was assessed using standardized mean differences (SMD), with SMD < 0.1 considered indicative of adequate balance.

Missing data

Complete data were available for > 95% of patients for all primary and secondary outcomes. For laboratory variables (lactate, pH, CRP), missingness was < 5% and determined to be missing completely at random (MCAR) based on Little’s test (p = 0.32). Multiple imputation with 10 imputed datasets using chained equations was performed for missing laboratory values, incorporating all covariates and outcomes in the imputation model. Sensitivity analyses comparing complete-case analysis and imputed results showed no meaningful differences; therefore, results from imputed analyses are presented.

Survival analysis

Survival analysis was performed using Kaplan-Meier curves with the Log-rank test in the matched cohort. A Cox proportional hazards model was used to calculate adjusted hazard ratios (aHR) for mortality, controlling for residual imbalance. A formal test for interaction between treatment modality (PD vs. PL) and etiology (SIP vs. NEC) was performed using Cox proportional hazards modeling.

Sensitivity analyses

1. Temporal bias: To account for potential temporal bias related to the institutional practice change after 2015, we performed two sensitivity analyses: (a) inclusion of calendar year as a covariate in the propensity score model and all regression analyses, and (b) a restricted analysis limited to the period following the practice change (2015–2023).

2. Feeding outcomes: To account for potential confounding by non-GI comorbidities that could independently delay feeding, we performed a sensitivity analysis excluding patients with: (a) severe intraventricular hemorrhage (Grade III-IV), (b) chronic lung disease (oxygen dependence at 36 weeks corrected gestational age), (c) hemodynamically significant PDA requiring medical or surgical treatment, and (d) culture-proven sepsis unrelated to abdominal source during the feeding advancement period.

Subgroup analyses

Subgroup analyses were pre-specified by etiology (NEC vs. SIP). A two-tailed p-value < 0.05 was considered significant. Analyses were performed using R software (version 4.3.1).

Results

Patient characteristics and propensity score matching

A total of 126 infants met inclusion criteria (PD-first = 66, PL-first = 60). Before matching, the PL-first group had higher birth weight (mean 920 g vs. 845 g, p = 0.04) and a greater proportion of NEC diagnoses (71.7% vs. 51.5%, p = 0.02). After PSM with 15 covariates, 47 matched pairs (n = 94) were generated with excellent balance across all covariates. Of the 94 matched infants, 69 (73.4%) were ELBW (< 1000 g) and 25 (26.6%) were VLBW (1000–1499 g), with similar distribution between treatment groups (Table 1).

Table 1.

Birth weight distribution in matched cohort (n = 94)

Birth Weight Category	PD-First (n = 47)	PL-First (n = 47)	Total (n = 94)
ELBW (< 1000 g), n (%)	35 (74.5%)	34 (72.3%)	69 (73.4%)
VLBW (1000–1499 g), n (%)	12 (25.5%)	13 (27.7%)	25 (26.6%)

ELBW Extremely Low Birth Weight, VLBW Very Low Birth Weight

Subgroup Analysis by Birth Weight Category: To determine whether treatment effects differed by birth weight, we performed a post-hoc subgroup analysis comparing ELBW (< 1000 g, n = 69) and VLBW (1000–1499 g, n = 25) infants. In the SIP cohort, the survival benefit of PD was observed in both weight categories (ELBW: HR 0.29, 95% CI 0.09–0.92; VLBW: HR 0.42, 95% CI 0.10–1.78), with no significant interaction (p = 0.62). In the NEC cohort, the trend favoring PL was similarly consistent across weight categories (p-for-interaction = 0.88). These findings suggest that the differential treatment effects by etiology are not driven by birth weight alone and support the application of our recommendations across the VLBW spectrum (Table 2).

Table 2.

Subgroup analysis by birth weight category (< 1000 g vs. 1000–1499 g)

Outcome	ELBW (< 1000 g)	VLBW (1000–1499 g)	p-for-interaction
SIP Cohort (n = 41)	(n = 30)	(n = 11)
- PD mortality	2/14 (14.3%)	1/5 (20.0%)	0.76
- PL mortality	8/16 (50.0%)	2/6 (33.3%)	0.48
- HR for PD (95% CI)	0.29 (0.09–0.92)	0.42 (0.10–1.78)	0.62
NEC Cohort (n = 53)	(n = 39)	(n = 14)
- PD mortality	12/21 (57.1%)	3/6 (50.0%)	0.74
- PL mortality	8/18 (44.4%)	2/8 (25.0%)	0.39
- HR for PD (95% CI)	1.48 (0.64–3.42)	1.62 (0.52–5.02)	0.88

*No statistically significant interaction between birth weight category and treatment effect in either etiology (p-for-interaction > 0.05 for all comparisons).*

Overall outcomes in matched cohort

In the matched cohort, overall 90-day mortality was 35.1% (33/94). There was no significant difference between groups: 16 deaths in the PD group (34.0%) and 17 in the PL group (36.2%) (p = 0.81). Kaplan-Meier survival curves were superimposable (Log-rank p = 0.79) (Fig. 1).

Fig. 1.

Kaplan-Meier Survival Curves for 90-Day Mortality (Overall Matched Cohort). *(Two overlapping survival curves (PD-first in blue, PL-first in red) showing nearly identical survival probability over 90 days. X-axis: Days post-procedure. Y-axis: Survival probability. Log-rank p = 0.79.)*

A formal test for interaction between treatment modality (PD vs. PL) and etiology (SIP vs. NEC) was statistically significant (p-for-interaction = 0.02), confirming that the effect of initial surgical strategy on 90-day mortality differs significantly between SIP and NEC subgroups. This finding provides statistical justification for our etiology-stratified analysis.

SIP cohort outcomes

Baseline characteristics of SIP patients

Of the 41 patients with confirmed SIP, 19 (46.3%) underwent PD-first and 22 (53.7%) underwent PL-first. Baseline characteristics were well-balanced between groups after propensity score matching.

Mortality in SIP

In the SIP cohort, PD-first was associated with significantly lower 90-day mortality compared to PL-first (15.8% vs. 45.5%, p = 0.03). The adjusted hazard ratio for mortality with PD was 0.32 (95% CI 0.11–0.91, p = 0.03). When analyzing attributable mortality (deaths directly related to intestinal pathology or surgical management), the difference became more pronounced: 5.3% (1/19) in the PD group versus 31.8% (7/22) in the PL group (p = 0.03).

Surgical morbidity in SIP

Major surgical complications were significantly less common in the PD-first group (21.1% vs. 54.5%, p = 0.03). This difference was driven primarily by lower rates of wound complications and stoma-related issues in the PD group. Notably, wound infections in the SIP-PL group were often associated with contamination from stoma output, while no wound complications occurred in the SIP-PD group.

Nutritional outcomes in SIP

Time to full enteral feeds was significantly shorter in the PD group (median 38 vs. 52 days, p = 0.04). Duration of PN showed a trend toward shorter duration in the PD group (35 vs. 48 days, p = 0.05). No cases of SBS occurred in the SIP-PD group, compared to 2 cases (9.1%) in the SIP-PL group (p = 0.31).

Long-term follow-up of SIP patients managed with PD alone

Among the 19 SIP patients initially managed with PD, 14 (73.7%) did not require subsequent laparotomy and were managed definitively with drainage alone. These patients were followed for a median of 24 months (IQR 12–36 months) post-discharge. Given the small sample size (n = 14) and that contrast studies were performed only for clinical indications (n = 6), these observations should be considered exploratory. No patient developed intestinal obstruction requiring surgical intervention. One patient (7.1%) developed symptoms consistent with intestinal dysmotility that resolved with medical management. No cases of late-onset intestinal stricture or stenosis were identified on the six contrast studies performed for clinical indications. Larger prospective studies with routine surveillance imaging are needed to definitively assess the risk of late complications (Table 3; Fig. 2).

Fig. 2.

Patient Flow Diagram (CONSORT-style). *(Flowchart showing: Assessed for eligibility (n = 145). Excluded (n = 19): Congenital anomalies (n = 8), comfort care (n = 11). Eligible (n = 126). Allocated to PD-first (n = 66). Allocated to PL-first (n = 60). Propensity Score Matching → Matched PD-first (n = 47), Matched PL-first (n = 47). Analyzed for outcomes (n = 94). Subgroups: SIP (n = 41), NEC (n = 53).)*

Table 3.

Comprehensive Outcomes in SIP Cohort by Treatment Group

Outcome	SIP-PD (n = 19)	SIP-PL (n = 22)	p-value
Mortality
90-day mortality, n (%)	3 (15.8%)	10 (45.5%)	0.03
Adjusted HR (95% CI)	0.32 (0.11–0.91)	Reference	0.03
Attributable mortality, n (%)	1 (5.3%)	7 (31.8%)	0.03
Surgical morbidity, N (%)	4 (21.1%)	12 (54.5%)	0.03
- Wound infection/dehiscence	0 (0%)	5 (22.7%)	0.04
- Stoma complication	0 (0%)	4 (18.2%)	0.08
- Intra-abdominal abscess	3 (15.8%)	2 (9.1%)	0.45
- Fascial dehiscence	1 (5.3%)	2 (9.1%)	0.57
Nutritional outcomes
- Time to full feeds (days), median (IQR)	38 (25–58)	52 (38–78)	0.04
- PN duration (days), median (IQR)	35 (22–60)	48 (35–72)	0.05
- Short Bowel Syndrome, n (%)	0 (0%)	2 (9.1%)	0.31
Reoperations
- Any reoperation, n (%)	5 (26.3%)	14 (63.6%)	0.02
-- Unplanned reoperations, n (%)	5 (26.3%)	3 (13.6%)	0.28
-- Planned stoma closure, n (%)	0 (0%)	11 (50.0%)	< 0.001

Data in bold are statistically significant

NEC cohort outcomes

Baseline characteristics of NEC patients

Of the 53 patients with confirmed NEC, 27 (50.9%) underwent PD-first and 26 (49.1%) underwent PL-first. Baseline characteristics were balanced after matching.

Mortality in NEC

In the NEC cohort, there was a non-significant trend toward lower mortality with PL-first (38.5% vs. 55.6%, p = 0.24). The adjusted hazard ratio for mortality with PD was 1.52 (95% CI 0.68–3.41, p = 0.31). However, the pattern of deaths differed significantly between groups, with progressive NEC accounting for 66.7% of deaths in the PD group compared to 0% in the PL group (p = 0.002). Attributable mortality was 44.4% (12/27) in the PD group versus 26.9% (7/26) in the PL group (p = 0.18).

Surgical morbidity in NEC

Major surgical complications were significantly more common in the PL-first group (57.7% vs. 29.6%, p = 0.04). However, the nature of complications differed: PL-group complications were predominantly wound-related (30.8%) and stoma-related (23.1%), while PD-group complications were primarily intra-abdominal abscesses (18.5%) requiring drainage.

Rescue laparotomy after PD in NEC

Among NEC patients initially managed with PD, 51.9% (14/27) required rescue laparotomy, typically performed for clinical deterioration within 72 h (median 48 h, IQR 24–72 h).

Nutritional outcomes in NEC

Time to full feeds showed a trend favoring PD (median 45 vs. 58 days, p = 0.08). SBS incidence was similar between groups (PD: 11.1% vs. PL: 7.7%, p = 0.54).

Causes of mortality

A detailed analysis of causes of death was performed for the 38 patients who died within 90 days (Table 4). Overall, the majority of deaths were directly attributable to complications of the underlying intestinal pathology or its management. Sepsis with multi-organ failure was the most common cause of death (21/38, 55.3%), followed by progressive NEC with intestinal necrosis (8/38, 21.1%). Importantly, the distribution of causes differed between treatment groups and etiologies. In the SIP subgroup, all deaths in the PD group (n = 3) were attributable to overwhelming sepsis, while in the PL group, deaths were distributed between postoperative complications (n = 4), sepsis (n = 3), and respiratory causes (n = 3). In the NEC subgroup, progressive disease was the dominant cause of death in the PD group (10/15, 66.7%), while postoperative complications and sepsis were more common in the PL group (Fig. 3).

Table 4.

Causes of 90-Day Mortality by Subgroup and Treatment

Cause of Death	SIP Subgroup (n = 13 deaths)		NEC Subgroup (n = 25 deaths)
	PD (n = 3)	PL (n = 10)	PD (n = 15)	PL (n = 10)
Sepsis/Multi-organ failure	3 (100%)	3 (30%)	9 (60%)	6 (60%)
Progressive NEC/SIP	0 (0%)	0 (0%)	10 (67%)*	0 (0%)*
Postoperative complications	0 (0%)	4 (40%)	2 (13%)	3 (30%)
Respiratory failure/BPD	0 (0%)	3 (30%)	2 (13%)	1 (10%)
IVH/Neurologic	0 (0%)	0 (0%)	1 (7%)	0 (0%)

BPD Bronchopulmonary dysplasia, IVH Intraventricular hemorrhage

Statistically significant difference between PD and PL in NEC subgroup (p = 0.002)*

Fig. 3.

Forest Plot of Adjusted Hazard Ratios for 90-Day Mortality by Subgroup. *(A forest plot with a vertical line at HR = 1.0. Three rows: 1) Overall Matched Cohort: HR 0.95 (CI 0.48–1.87). 2) SIP Subgroup: HR 0.32 (CI 0.11–0.91). 3) NEC Subgroup: HR 1.52 (CI 0.68–3.41). The SIP subgroup diamond is significantly to the left of the line.)*

Sensitivity analyses

Temporal bias analysis

To address potential era effects related to the institutional practice change after 2015, we performed a sensitivity analysis restricted to the period 2015–2023 (n = 89). The results were consistent with the main analysis: overall mortality remained comparable between PD-first (32.4%) and PL-first (34.1%) groups (p = 0.86). In the SIP subgroup (n = 32), PD continued to demonstrate significantly lower mortality (16.7% vs. 46.2%, p = 0.04; HR 0.30, 95% CI 0.10–0.89). In the NEC subgroup (n = 57), the mortality trend favoring PL persisted (PD: 52.4% vs. PL: 37.5%, p = 0.28).

Sensitivity analysis for time to full feeds

To determine whether differences in time to full feeds were confounded by non-GI comorbidities, we performed a sensitivity analysis excluding 28 patients with significant comorbidities that could independently delay feeding (severe IVH: n = 12, CLD: n = 18, significant PDA: n = 15, non-GI sepsis: n = 20; some patients had multiple comorbidities). In this restricted cohort (n = 66), the difference in time to full feeds persisted and remained statistically significant: PD-first group median 40 days (IQR 26–62) vs. PL-first group median 52 days (IQR 34–78), p = 0.03. Multivariable regression adjusting for the presence of these comorbidities confirmed that PD-first was independently associated with shorter time to full feeds (β coefficient − 8.5 days, 95% CI -15.2 to -1.8, p = 0.02) (Fig. 4).

Fig. 4.

Comparative Bar Charts of Key Outcomes by Etiology and Treatment Group. (Four grouped bar charts: A) Mortality: SIP (PD bar low, PL bar high), NEC (bars closer). B Surgical Morbidity: SIP (PD bar low, PL bar high), NEC (PD bar moderate, PL bar high). C Rescue Laparotomy Rate: SIP (low bar), NEC (high bar). D Time to Full Feeds: SIP (PD short bar, PL long bar), NEC (similar trend).)

Sensitivity analysis for etiology misclassification

To address potential misclassification of etiology (NEC vs. SIP) based on imaging alone, we performed a restricted analysis limited to 78 patients who had surgical confirmation of diagnosis (intraoperative findings documented in operative reports). The results were consistent with the main analysis: In the SIP subgroup (surgically confirmed, n = 32), PD was associated with lower mortality (16.0% vs. 42.9%, HR 0.35, 95% CI 0.12–0.99). In the NEC subgroup (surgically confirmed, n = 46), the mortality trend favoring PL persisted (PD: 52.2% vs. PL: 39.1%, HR 1.48, 95% CI 0.62–3.54). These findings suggest that our main results are robust to potential diagnostic misclassification.

As-treated analysis addressing crossover bias

In our primary analysis, we used intention-to-treat (ITT) grouping based on the initial procedure. However, 19 patients (40.4%) in the PD-first group subsequently underwent rescue laparotomy (crossover). To address potential crossover bias, we performed an as-treated analysis classifying patients according to the final surgical management (PD alone vs. laparotomy at any time). The results were consistent with the ITT analysis: In the SIP subgroup, patients managed with PD alone (n = 14) had lower mortality compared to those undergoing laparotomy at any time (n = 27) (14.3% vs. 40.7%, HR 0.33, 95% CI 0.10–1.02). In the NEC subgroup, patients undergoing laparotomy at any time (n = 33) had lower mortality compared to those managed with PD alone (n = 20) (39.4% vs. 60.0%, HR 1.55, 95% CI 0.71–3.38). These findings confirm that our ITT results are not substantially biased by crossover .

Discussion

This detailed retrospective cohort study, utilizing comprehensive propensity score methodology with 15 illness severity covariates, sensitivity analyses for temporal bias and confounding, formal interaction testing, and long-term follow-up data, provides compelling evidence that the optimal initial surgical intervention for intestinal perforation in ELBW infants is fundamentally dependent on the underlying disease etiology. Our findings reconcile some of the contradictions in the existing literature by demonstrating that the aggregated results of prior RCTs may mask opposing treatment effects in NEC and SIP subgroups.

SIP: the case for peritoneal drainage as the intervention of choice

The significant mortality reduction with PD for SIP (aHR 0.32; attributable mortality 5.3% vs. 31.8%) is the most salient finding of this study. This aligns with the pathophysiological profile of SIP—a focal defect in an otherwise healthy bowel without extensive necrosis or pan-intestinal inflammation [9]. PD effectively controls peritoneal soilage and sepsis with minimal physiological insult, avoiding the stress of laparotomy, anesthesia, and potential fluid/heat losses in a tenuous infant. Our high complication rate with PL for SIP (54.5%) underscores the disproportionate risk of a major operation for a focal problem.

An important concern regarding the use of PD as potentially definitive management in selected cases for SIP is the potential for late complications, particularly intestinal stricture or obstruction. Our long-term follow-up data (median 24 months) are reassuring in this regard, with no cases of surgically significant obstruction or stricture identified among the 14 SIP patients managed with PD alone. This finding aligns with the pathophysiological understanding of SIP as a focal perforation without the diffuse ischemic insult characteristic of NEC [9] and supports the hypothesis put forward by Bucher et al. [10] that PD may be potentially definitive management in selected cases for most SIP cases [11, 12].

NEC: the persistent dilemma and the role of laparotomy

For NEC, our data reflect the complexity captured in the NECSTEPS trial [5]. The non-significant mortality trend favoring PL (38.5% vs. 55.6%), coupled with a > 50% rescue laparotomy rate after PD, suggests that for diffuse necrotizing disease, PD often functions as a temporizing measure rather than potentially definitive management in selected cases. The striking difference in causes of death—progressive NEC accounting for 67% of PD deaths versus 0% of PL deaths—underscores the importance of definitive source control through resection of necrotic bowel.

However, this survival benefit comes at a cost, as evidenced by our significantly higher surgical morbidity in the PL group (57.7% vs. 29.6%). This trade-off—potential survival benefit versus increased immediate surgical risk—explains the ongoing clinical equipoise. Our findings are congruent with a meta-analysis by Rao et al. [6], which concluded that while PD may stabilize the sickest infants, it does not reduce mortality or SBS in NEC compared to PL.

Nutritional and morbidity outcomes

The shorter time to full enteral feeds in the PD group is a novel finding that persisted after rigorous sensitivity analyses. Regarding long-term outcomes in SIP patients managed with PD alone, our data (n = 14) are reassuring but must be interpreted with caution due to the small sample size and lack of routine surveillance imaging. No cases of intestinal obstruction were observed, but we cannot exclude a low incidence of stricture or dysmotility. Future prospective studies with standardized long-term follow-up protocols are needed.

The distinction between planned and unplanned reoperations (Table 5) provides a fairer comparison between groups. While PL was associated with more total operations, these were predominantly planned stoma closures, whereas unplanned reoperations for complications were actually more common in the PD group (51.9% vs. 7.7% in NEC; 26.3% vs. 13.6% in SIP). This nuance is critical for clinical decision-making and family counseling (Table 6).

Table 5.

Comprehensive Outcomes in NEC Cohort by Treatment Group

Outcome	NEC-PD (n = 27)	NEC-PL (n = 26)	p-value
Mortality
90-day mortality, n (%)	15 (55.6%)	10 (38.5%)	0.24
Adjusted HR (95% CI)	1.52 (0.68–3.41)	Reference	0.31
Attributable mortality, n (%)	12 (44.4%)	7 (26.9%)	0.18
Surgical morbidity, N (%)	8 (29.6%)	15 (57.7%)	0.04
- Wound infection/dehiscence	4 (14.8%)	8 (30.8%)	0.16
- Stoma complication	2 (7.4%)	6 (23.1%)	0.11
- Intra-abdominal abscess	5 (18.5%)	3 (11.5%)	0.44
- Fascial dehiscence	1 (3.7%)	2 (7.7%)	0.48
Nutritional outcomes
- Time to full feeds (days), median (IQR)	45 (32–72)	58 (40–92)	0.08
- PN duration (days), median (IQR)	42 (28–75)	55 (38–95)	0.09
- Short Bowel Syndrome, n (%)	3 (11.1%)	2 (7.7%)	0.54
Reoperations
- Any reoperation, n (%)	14 (51.9%)	17 (65.4%)	0.32
-- Unplanned reoperations, n (%)	14 (51.9%)	2 (7.7%)	< 0.001
-- Planned stoma closure, n (%)	0 (0%)	15 (57.7%)	< 0.001
- Rescue laparotomy after PD, n (%)	14 (51.9%)	N/A	-

Data in bold are statistically significant

Table 6.

Summary of Key Literature Compared to Present Study Findings

Study (Year)	Design	Key Finding (Overall)	Comment on Etiology Subgroup
Moss et al. (NECSTEPS, 2006) [5]	RCT (n = 117)	No diff in 90-day mortality (PD 34.5% vs. PL 35.5%)	Post-hoc analysis suggested possible benefit of PL in NEC with >1 kg BW; not powered for subgroup
Blakely et al. (NICHD, 2006) [4]	RCT (n = 69)	No diff in survival at 90 days or neurodevelopment	Trend towards more reoperations in PD group, consistent with our NEC findings
Rao et al. Cochrane (2011) [6]	Meta-analysis	PD does not reduce mortality or SBS vs. PL	Highlighted significant heterogeneity; subgroup analysis by etiology not performed
Present Study (2024)	Retrospective Cohort (PSM)	No overall mortality difference; p-interaction = 0.02	Clear divergence: PD superior in SIP (HR 0.32); PL favored in NEC with trend (HR 1.52)

Limitations

While our study has several methodological strengths, including comprehensive propensity score adjustment with 15 covariates, several limitations warrant careful consideration.

Propensity score limitations

Despite inclusion of 15 clinically relevant covariates, PSM can only adjust for measured confounders. Unmeasured or incompletely measured confounders—such as subtle differences in clinical judgment (e.g., perceived futility, family preferences), the exact timing of perforation relative to presentation, or nuanced variations in bedside examination findings—may have influenced treatment allocation. Additionally, the quality of resuscitation prior to surgical intervention (e.g., fluid bolus administration, timing of antibiotic therapy) was not uniformly recorded and may represent residual confounding.

Unmeasured confounders

Specific variables that could influence outcomes but were not available in our dataset include: detailed echocardiographic assessment of pulmonary hypertension, severity of pre-existing chronic lung disease, and baseline neurodevelopmental status prior to perforation. These factors may independently affect both treatment decisions and mortality.

Single-center design

While ensuring consistency in perioperative and postoperative care protocols, our single-center design limits generalizability. Practice patterns, surgeon expertise, and patient case mix may differ across institutions, and our findings require external validation.

Sample size considerations

The modest sample size, particularly in subgroup analyses (SIP-PD n = 19; SIP-PD long-term follow-up n = 14), limits the precision of our estimates. Confidence intervals for some comparisons (e.g., NEC mortality HR 1.52, 95% CI 0.68–3.41) include unity, and we cannot exclude the possibility of no true difference. The interaction test (p = 0.02) should be interpreted cautiously given the sample size, and we emphasize that this finding requires confirmation in larger prospective cohorts.

Verification bias

As noted in “Long-term follow-up of SIP patients managed with PD alone” section, contrast studies in long-term follow-up were performed only for clinical indications. Thus, the absence of identified strictures or stenosis may underestimate true prevalence.

Interaction testing considerations

The significant interaction between treatment modality and etiology (p = 0.02) provides statistical support for our stratified analysis. However, given the modest sample size (n = 94 overall, n = 41 SIP, n = 53 NEC), the interaction estimate has wide confidence intervals. We performed a post-hoc power calculation: with the observed effect sizes, the power to detect the interaction at α = 0.05 was approximately 65%. Therefore, while the interaction is statistically significant, it should be interpreted as hypothesis-generating rather than definitive. Independent validation in larger cohorts is required.

Clinical implications and future directions

This study argues decisively for an etiology-driven algorithm (Fig. 5). When SIP is suspected based on the absence of pneumatosis and presence of risk factors (early postnatal steroids, indomethacin exposure), PD may be considered as the initial intervention. Importantly, PD failure requiring rescue laparotomy was defined by specific clinical triggers within 72 h (persistent shock, rising lactate, worsening acidosis). In our series, 26% of SIP patients required rescue laparotomy, while 74% were managed definitively with PD alone. For classic NEC with pneumatosis, primary laparotomy should be strongly considered, especially if the infant demonstrates any physiological reserve, recognizing that while surgical morbidity is higher, this approach may offer a observed survival association through definitive source control. In the unstable NEC patient with no physiological reserve, PD may serve as a bridging therapy to allow stabilization before delayed laparotomy.

Fig. 5.

Proposed Etiology-Driven Clinical Algorithm for Surgical Management

Future research should focus on: (1) refining preoperative diagnostic tools to distinguish NEC from SIP more accurately (e.g., novel biomarkers, advanced imaging); (2) prospective trials randomizing infants within these etiological subgroups; (3) longer-term follow-up studies assessing neurodevelopmental and gastrointestinal outcomes through childhood; and (4) quality improvement initiatives to reduce surgical morbidity in laparotomy patients (e.g., enhanced wound closure techniques, standardized stoma management protocols).

• Step 1: Preterm infant with pneumoperitoneum (perforation already confirmed radiographically).

• Step 2: Imaging assessment for pneumatosis intestinalis or portal venous gas:

NO (Suspected SIP) → Proceed to Primary Peritoneal Drainage (bedside, under local anesthesia).

If clinical deterioration persists/occurs despite PD (defined as persistent shock, rising lactate, worsening acidosis within 72 h) → Rescue Laparotomy.

If stable → Continue conservative management (PD as potentially definitive management in selected cases).

YES (Suspected NEC) → Assess Physiological Reserve (hemodynamic stability, ventilatory requirements, inotrope dependence).

Unstable / No reserve → Consider PD as bridging therapy (temporizing measure).

Any reserve → Recommend Primary Laparotomy (definitive source control).

Conclusion

In preterm infants with intestinal perforation, peritoneal drainage and primary laparotomy are not universally equivalent options. PD is associated with lower observed mortality and lower morbidity in spontaneous intestinal perforation, with reassuring long-term outcomes, likely representing definitive management for most cases. For necrotizing enterocolitis, primary laparotomy may offer a mortality advantage despite higher immediate surgical morbidity, as it directly addresses the diffuse necrotic process. Surgical decision-making must therefore be guided by the best available evidence regarding underlying etiology, moving beyond a one-size-fits-all approach to a precision-based strategy in this vulnerable population.

Authors’ contributions

M.M.S.: Conceptualization, Methodology, Formal Analysis, Writing – Original Draft- A.R.N.: Data Curation, Investigation, Resources, Writing – Review & Editing- I.A.M.: Methodology, Validation, Formal Analysis, Visualization- M.S.: Investigation, Resources, Data Curation, Project Administration- M.A.: Supervision, Validation, Writing – Review & Editing, Final ApprovalAll authors have read and agreed to the published version of the manuscript.

Funding

The authors received no specific funding for this work.

Data availability

The data underlying this article will be shared on a reasonable request to the corresponding author, pending approval from the Institutional Review Board.

Declarations

Ethics approval and consent to participate

This retrospective cohort study was approved by the Institutional Review Board (IRB) / Research Ethics Committee (REC) of our Teaching Tertiary Hospital. All procedures performed were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Due to the retrospective nature of the study, the need for individual informed consent was waived by the approving ethics committee of our Teaching Tertiary Hospital. Patient data were anonymized and de-identified prior to analysis to ensure confidentiality.

Consent for publication

Not applicable. No individual personal data (including details, images, or videos) are presented in this manuscript.

Competing interests

The authors declare no competing interests.