Phototherapy: a new risk factor for necrotizing enterocolitis in very low birth weight preterm infants? a retrospective case–control study

To investigate the association between phototherapy (PT) and the development of necrotizing enterocolitis (NEC) in very low birth weight (VLBW) infants. A retrospective case–control study was conducted on VLBW infants with or without NEC (stage IIA or greater) born at ≤35 weeks’ gestation in a tertiary hospital over 7 years. Sample size calculation, trend test, as well as univariate and multiple logistic regression analyses were employed. A total of 824 VLBW infants were reviewed, with 74 cases and 122 controls finally enrolled. The odds of NEC increased with the duration and number of PT sessions. Exposure to >120 h and >4 instances of PT were significantly associated with NEC in multivariate analysis. This is the first study suggesting a potential association between PT and development of NEC in VLBW infants. This association needs further exploration.


INTRODUCTION
Necrotizing enterocolitis (NEC) is a gastrointestinal emergency and one of the most devastating diseases encountered in neonates, with high morbidity and mortality rates [1].Although the etiology of NEC remains unclear, its pathophysiology is generally hypothesized to be multifactorial [2,3].A combination of genetic predisposition, intestinal immaturity, imbalance in microvascular tone, abnormal microbial colonization in the intestine, and a highly immunoreactive intestinal mucosa leads to a confluence of predisposing factors [4].
Phototherapy (PT) is the standard treatment for neonatal jaundice, used to reduce unconjugated bilirubin [5], and is commonly administered in neonatal intensive care units.PT reduces bilirubin levels through three possible mechanisms: structural isomerization to lumirubin, photoisomerization to a less toxic bilirubin isomer, and photo-oxidation to polar molecules [6].It is generally considered safe [7].However, the long-standing assumption that this therapy is innocuous with no serious adverse effects for even the most immature babies is neither evidence based nor biologically plausible [8].A 2022 study by Sainan Fan on neonatal stools suggested that PT can significantly affect the intestinal probiotic flora of newborns, particularly reducing Bifidobacterium, which plays a crucial role in protecting against NEC [9].Inappropriate colonization in preterm infants leads to increased inflammatory responses and abnormal bacterial glycosylation patterns [3,4].Furthermore, some studies have linked PT to altered nitric oxide and endothelin-1 levels and increases in proinflammatory cytokines such as interleukin (IL) −2, IL-8 and IL-10 [10,11].The increase in IL-8 and the excessive inflammatory response has been observed to be involved in the development of NEC [4].The excessive immature inflammatory response associated with abnormal intestinal microbiota is currently considered the most likely basis for the pathogenesis of NEC [4].Moreover, numerous studies have shown that irradiance and infrared radiation from PT affect mesenteric and peripheral blood flow in newborns, which may indicate mesenteric ischemia [10][11][12].Hypoxia and ischemia lead to imbalance in microvascular tone by producing vascular regulators such as nitric oxide and endothelin [4] and shift cells to oxidative stress by producing free radicals [13].Based on the above literature, we hypothesize that PT might increase the risk of NEC in VLBW infants.
To our knowledge, no clinical studies have systematically investigated the possible association between PT and the development of NEC in VLBW neonates.This is of considerable concern due to the general increase in irradiance of PT lamps [14].Therefore, we conducted this study with the aim of investigating whether there is an association between PT and NEC in VLBW infants.
Hospital of Chongqing Medical University.Our neonatal diagnostic center serves approximately 5000 neonates annually.This study adhered to the Observational Studies in Epidemiology (STROBE) reporting guideline.
VLBW infants born before 35 weeks of gestation were included.Infants were divided into NEC and non-NEC groups, which served as the case and control groups respectively.We aimed to match each NEC case during the same study year for gender, gestational age (difference ≤1 week), birth weight (difference ≤150 g), small for gestational age (SGA) status, and mode of delivery to two controls.Exclusion criteria included maternal chorioamnionitis, early onset sepsis, severe congenital anomalies, and incomplete information.Spontaneous intestinal perforations were also excluded, as they likely represented a different disease entity with distinct pathogenesis [4].For the control group, we excluded infants who died before the postnatal age of NEC onset in the corresponding case.

Data collection
Information on observation factors were obtained from the electronic medical record system of the Women and Children's Hospital of Chongqing Medical University and anonymized before analysis.Data were collected for each case and control from birth up to the day before the documented NEC onset in the cases and the corresponding controls.Demographic characteristics, perinatal features, treatment, laboratory tests, neonatal comorbidities, and information on surgical procedures were reviewed.Feeding and feeding speed were obtained from nursing documentation.The medical Ethical Committee approved this retrospective research study (N2022.031)and waived the written informed consent.

Treatment
The decision to use PT was made according to total serum bilirubin (TSB) values recommended by the Chinese Pediatric Society guidelines for neonates born at less than 35 weeks' gestation, issued in 2001 (Supplementary Table 1) [5].Guidelines did not change during the study.Recommendations for PT were based on bilirubin concentration, postnatal age, birth weight and gestational age.In our unit, PT may also be initiated at a TSB 2 mg/dl below the regular level for infants with neurotoxic risk factors, including acidosis (pH ≤ 7.2), sepsis, asphyxia (5-min Apgar score <5), and hemolytic disease [15].PT was terminated when either the TSB or transcutaneous bilirubin (TCB) decreased 2 mg/dl below the cutoff for PT.PT was provided with blue light-emitting diode (LED) PT system (Zhengzhou Dison Instrument and Meter Co., Ltd, Henan, China or Ningbo David Medical Device Co., Ltd, Zhejiang, China, spectrum range 400-550 nm).The PT lamp irradiance intensities were preadjusted so that typical patient positioning would result in delivery of approximately 8 ~10μW/cm2 /nm.In treating each infant, the lamp was positioned to deliver the target irradiance, with genitals and eyes covered.Exchange transfusion was considered when TSB reached the exchange transfusion cutoff value.During the study period, infants had regular TCB (every 8-12 h) checks.TCB measurements were performed by a twin-beam microbilimeter (Ningbo David Medical Device Co., Ltd, Zhejiang, China).TSB was collected at the beginning of the first PT session and when the TCB level plus 2 mg/dl reached the threshold recommended for PT.For each newborn, the first TSB collected was considered the qualifying TSB.Consequently, all the initial TSBs were obtained before PT.

Definitions
The duration and number of PT courses were defined as the total number of hours and instances an infant received PT before the diagnosis of NEC, respectively.NEC was diagnosed using the accepted combination of clinical signs and radiographic findings via the Bell's Modified Staging Criteria, and infants with NEC (Bell's stage ≥ II) were included in our study [3].Hemolytic disease of the newborn was confirmed by a positive direct antiglobulin test and/or positive antibody release test, indicating antibodies on red blood cell (RBC) [2].Patent ductus arteriosus (PDA) was diagnosed based on echocardiographic findings.SGA was defined as a birth weight below the 10th percentile for gestational age and gender [16].Adequate antenatal steroids were defined as 6 mg intramuscular dexamethasone injections given every 12 h for four doses within 2-7 days before delivery [17].

Potential confounding variables and statistical analysis
To investigate confounding by indication, we included indicator variables as categorical variables for the difference between the first TSB level and the threshold for their risk group in 1-mg/dL intervals [18] (e.g., 0-1.0 below mg/dL, 0-1.0 above mg/dL, etc.), and hemoglobin (Hb) at admission was categorized in 20 g/L intervals (e.g., <130 g/L, 130 to <150 g/L) [2].Due to the strong association between PT duration and number (the correlation coefficient was 0.877 by Spearman's rank correlation test), they were analyzed with separate models and coded in 30-hour and 2-category intervals, respectively (e.g., 0-≤30 h; 0-≤2, etc.).The 10th percentile (approximately 30 h and 2 PT sessions) was used as the reference.Maximum TSB level was not collected as a candidate predictor since it could be affected by PT, potentially acting as a mediator.
Univariate analyses compared cases and controls.Categorical variables were analyzed with chi-square tests.Continuous variables were analyzed using Student's t test or the Mann-Whitney U test, depending on data distribution.Multivariable logistic regression adjusted for potential confounding variables.Variables entered in the logistic regression model included covariables and potential confounding variables.Covariables were variables likely related to NEC at p < 0.1 in univariate analysis, whereas confounder selection was illustrated in the directed acyclic graph (DAG).Tests for linear trend were performed by entering the median value of each PT duration and number category as continuous variables in the models.To assess NEC influence associated with PT by TSB, separate regression analyses were performed, with the difference between the first TSB and threshold as well as PT added as predictor variables in different models of duration and number.Odds ratios (ORs) were used to compare crude and adjusted results.Statistical significance was set at p < 0.05 by two-sided.Statistical analysis was conducted using SPSS V26.0 (IBM, Armonk, NY, USA).
Based on our preliminary study, a 1:2 matching control study in NEC (n = 30) and non-NEC (n = 60) groups of VLBW infants, with PT duration as the primary outcome, the difference between the groups means was estimated to be 20 h.According to the sample size calculation formula, 65 patients would need to be recruited for the NEC group with an alpha level of 0.05 and powered at 90%.

RESULTS
During the study period, 824 VLBW infants born at no more than 35 weeks of gestation were admitted to our unit.Of these, 94 (11.4%) were diagnosed with NEC, with 74 (9.0%) finally included in the study.The remaining 20 were excluded due to maternal chorioamnionitis (n = 3), early-onset sepsis (n = 5), gastrointestinal malformations (n = 8) including 7 spontaneous intestinal perforations and 1 congenital megacolon, and major congenital anomalies (n = 4) including 1 aortic constriction, 1 intraventricular hemorrhage (grade IV), 1 severe hydrocephalus, and 1 Prader-Willi Syndrome.Of the 74 enrolled cases, 36 (48.6%) required surgery, and 5 (6.8%) died without operation.The average time to NEC onset from birth was 13.2 ± 8.2 days.We aimed to match each case with two controls, resulting in 122 infants without NEC finally paired to the cases, with 26 cases matched to only one control.None of the cases and controls received exchange transfusion or tense PT.
As shown in Table 1 (see Supplementary Table 2 for more details), demographic characteristics, perinatal features, mechanical ventilation, antibiotics, umbilical catheters, and feeding speed were similar between cases and controls.In cases, antenatal steroids deficiency, formula feeding, and RBC transfusion were more frequent, and non-invasive continuous positive airway pressure (CPAP) therapy was shorter compared to controls and were enrolled as covariates (p < 0.1).Although not significant (p > 0.05), decrease Hb and reduced breast milk feeding increased the odds of NEC, while adequate antenatal steroids decreased the odds of NEC.Albeit not significantly, stratification of the difference between TSB and threshold showed that 1-2 above decreased the OR of NEC and 0-1 above increased the OR of NEC.No significant differences were observed between years of birth associated with NEC (Supplementary Table 3).Between 2019 and 2020, our unit collaborated with German specialists, leading to an increased admission of infants with younger gestational age.
All infants received PT, with the mean time to first PT from birth being 19.2 ± 8.8 h for cases and 18.4 ± 7.7 h for controls, showing no significant differences.In the univariate analysis (Table 2), the increase in PT duration had higher odds associated with NEC (trend P = 0.010), and this association was strengthened after adjustment (trend P = 0.013).Consistent with duration, the odds of NEC increased with increasing PT numbers (trend P = 0.033), and the aOR was strengthened after adjustment (trend P = 0.048).
Exposure to >120 h and >4 numbers of PT courses were both significantly associated with NEC in both univariate and multivariate analyses (p < 0.05).In additional analyses adjusted for TSB difference with different models of duration and number of PT respectively (Supplementary Table 4), in both models, the odds of  NEC increased with PT and decreased with a TSB difference above 1-2 mg/dL compared to that in univariate analyses.Subgroups analysis showed the distribution of the duration and number of PT for different NEC stages in the group of 74 cases (Table 3).There were no statistical differences in the duration and number of PT courses between stage II and stage III groups, and between death and non-death groups (p > 0.05).Of the 5 total deaths, 3 occurred in the group with >120 h of PT and 3 in the group with >4 numbers of PT, each accounting for a high proportion of 60%.

DISCUSSION
To the best of our knowledge, this is the first study suggesting that VLBW infants with increasing PT exposure tend to have a greater risk of NEC development, and exposure to >120 h and >4 numbers of PT courses are significantly associated with NEC.
We believe our findings are significant for three main reasons.(1) Learning from history, PT is a technique commonly used in neonates to prevent or treat high TSB, but the safety of PT in VLBW infants remains unanswered.The increased odds of the serious complication of NEC, with or without significant p values, need to be considered [10].(2) VLBW infants have long hospital stays.In this study, 14.9% and 43.2% of the NEC group had more than 120 h of PT and more than 4 numbers of PT courses, respectively (compared to 4.9% and 29.5% of the control group).
(3) Bradford Hill suggested a set of criteria to judge causality [19].Our study met several of these criteria.The association of NEC and PT appeared to be dose-related (biological gradient).PT, a type of ionizing radiation, is directly linked to the risk of radiation-induced enterocolitis (analogy) [20].The results were supported by existing evidence and they are biologically plausible.First, they could be supported by changes in the intestinal probiotic flora and increases in proinflammatory cytokines of newborns during PT [9][10][11], which is an important risk factor for NEC [4].Second, the increased cutaneous and peripheral blood flow contributed to PDA opening, leading to subsequent changes in mesenteric blood flow and associated mesenteric ischemia [6,12,21].Hypoxia, ischemia and the associated imbalance in microvascular tone may play a downstream role in the pathogenic cascade leading to NEC [4].The ischemic conditions contribute to oxidative insults, causing variable degrees of endothelial dysfunction and intestinal damage [4].Moreover, phototoxicity might directly result in oxidative injury to cell membranes by producing chemically reactive free radicals and specifically singlet and triplet forms of oxygen [10,14].VLBW infants, whose skin readily transmits light, would be most vulnerable to oxidative injury [7,10].Smaller infants also have lower hemoglobin and hematocrit (thus less light-absorbing hemoglobin), allowing more light to interact not only with bilirubin but also with molecules in many other tissues inside the body [22].Meanwhile, animal experiments documented that PT induced vascular smooth muscle relaxation possibly either by nitric oxide-cyclic guanosine monophosphate pathway or direct photo relaxation and induced apoptosis in neonatal small intestine [23,24].
In the study, we observed that increasing PT exposure increased the risk of NEC.In the 2008 Network trial comparing PT administered at different TSB levels for death or neurodevelopmental impairment, their secondary outcomes suggested no significant association between aggressive PT and NEC with a relative risk (RR) of 0.90 (95% CI [0.70-1.14])[25].However, their starting TSB standards were different, and moderate bilirubin levels may have clinically important antioxidant benefits [8,10,25].This could be a confounder, biasing outcomes, while the cutoff value remains unknown.In our additional analyses adjusted for the difference between first TSB and PT threshold with different models of duration and number of PT, the odds of NEC associated with PT increased, while the odds of NEC associated with TSB difference above 1-2 mg/dL decreased, partially corroborating this view.Furthermore, the primary risk factors contributing to the development of TSB during hospitalization include HDN, gestational age, bruising, and feeding.After adjusting for these variables, the odds of NEC occurrence increased.The irradiation intensity in the Network trial (mean irradiance levels 22-23 μW/cm 2 /nm) was greater than that in our study (8-10 μW/cm 2 /nm).The effect of PT on newborns, especially small infants, varies greatly with different irradiation levels, but the appropriate irradiance levels and their optimal application remain unclear.The Network trial did not stratify PT duration or provide the number of PT courses, preventing further comparing.We did not consider PT to be associated with NEC severity, as there was no trend change of odds and no significant difference in PT duration and number between stage II and stage III of NEC in our study.However, based on our small sample size of cases, further research is needed.
It is noteworthy that a high percentage of >120 h (60%) and >4 numbers (60%) of PT were found in the death group without significant difference (p > 0.05).This corroborates previous studies [8,25,26].The only large randomized clinical trial about cycled PT (≥15 min/h) vs continuous PT for controlling TSB among ELBW infants reported the adjusted point estimates for relative risk of death with cycled PT (RR 0.79; 95% CI, 0.40-1.54)[8].In the only other large Network trial, the findings for the relative risk of death with aggressive PT was 1.05 (95% CI, 0.90-1.22)among ELBW infants [25].These findings have been largely ignored due to their lack of significant at P < 0.05, an error often made in failing to recognize and seriously consider potential treatment hazards when statistical power is limited [14].The findings are biologically plausible [25].Higher mortality might be caused by oxidative damage to cell membranes due to lower antioxidant levels from lower bilirubin concentrations or by oxidative stress produced by PT, especially in the smallest, most immature infants [25].It is also possible that these infants are exposed to greater circulatory changes or instability caused by more prolonged PT [11].In our study, further studies were hampered by the small number of deaths in the case group and the wide confidence interval, which could not exclude serendipity.
The strength of our study lies in its being the first 7-year clinical observational investigation of the association between PT and NEC development in VLBW infants.Our single center study may have reduced inter-hospital variation in PT, although interoperator variation remains possible.Although retrospective, rigorous study designs included controlling confounding factors to minimize bias and sample size calculation.Our results reiterate caution regarding the potential harms of PT and the prudence of avoiding unnecessary PT [11], providing a reference future well-designed multicenter prospective studies.
Our study has several limitations.First, as a single center, nonrandomized, retrospective study, the generalizability of the results is limited.Second, the relatively small sample of 74 cases diagnosed with NEC, although deemed adequate in sample size calculation, raises the possibility of a type II error.Third, the criteria for PT application were based on factors such as gestational age, birth weight, days post-birth, and neurotoxicity, which might overlap with NEC risk factors, introducing potential bias.This possibility remains despite our attempts to mitigate this bias through matched pairs, univariate, and multivariate regression analyses.Last, given the multifactorial pathogenesis of the illness, the list of confounders is expected to be long, potentially influencing the findings of an observational study.
Our study suggests that increased PT exposure may raise the risk of NEC development in VLBW infants.In our results, a higher percentage of prolonged PT was found in the death group.The findings highlight the importance of neonatologists exercising caution when using PT in VLBW, considering the potential harms of NEC.

Table 1 .
Variables between cases and controls by univariate analysis.

Table 2 .
Association between phototherapy and NEC.

Table 3 .
Subgroup analysis between phototherapy and different stage of cases.