In this report, we describe the prevalence, risk factors, and outcomes associated with cholestasis in the setting of medical and surgical NEC. Supporting our hypothesis, irrespective of the NEC cohort, infants with cholestasis had worse outcomes compared to those without cholestasis. Previous descriptions of cholestasis in neonatal gastrointestinal conditions were made in heterogenous populations of infants with intestinal failure including surgical NEC along with numerous other diagnoses. Descriptions of cholestasis in medical NEC are sparse, with descriptions including patients with combined medical and surgical NEC populations (10). Our description of cholestasis in infants with NEC, and in particular with medical NEC, addresses this knowledge gap.
In our study, 38% of medical NEC and 71% of surgical NEC infants developed cholestasis. In a retrospective study spanning 10 years involving 178 patients, Veenstra et al. demonstrated a 54% prevalence of parenteral nutrition-associated cholestasis (PNAC) in infants with NEC (10). However, their study did not delineate between infants with medical and surgical NEC. Also, the focus of their study was PNAC, while ours was cholestasis alone irrespective of the duration of PN. PNAC is a form of cholestasis observed with a need for PN therapy longer than 14 days (12). Our approach differentiates between medical and surgical NEC patients, includes cholestasis where factors other than PN have played a role and links cholestasis to the NEC diagnosis by focusing on the peak bilirubin levels after NEC diagnosis.
We observed that infants of lower gestational age and lower birthweight had a higher prevalence of cholestasis. These findings are similar those by previous groups which have highlighted the increased incidence of NEC in more preterm infants (6, 13). Christensen et al. showed that the risk for PNAC increased with decreasing birth weight. Infants with birth weights of 1000-1499 g had an odds ratio estimate of 2.8 of developing PNAC, while in infants with birth weight <500g the odds increased to 30.7 (6). The median postmenstrual age (PMA) at NEC diagnosis in our cohort was 30.3 – 34.8 weeks, with those patients with surgical NEC being of younger PMA. In a population-based study, Llanos et al, demonstrated the peak incidence of NEC at a PMA of 32 – 35 weeks (14). Others have shown a bimodal distribution of NEC, again with those with younger birth gestational age being diagnosed with NEC later (15). However, both of these studies were representative of all NEC and did not differentiate the PMA of infants with medical NEC and surgical NEC.
Peak bilirubin between the medical NEC and surgical NEC groups was similar. However, in infants with medical NEC, the peak bilirubin occurred earlier than in infants with surgical NEC. The later peak of bilirubin in surgical NEC is most likely due to the prolonged course of cholestasis in these infants, as we were unable to determine the timing of cholestasis onset. Inflammation is a major inciting factor in the pathogenesis of cholestasis (16). Infants with surgical NEC have recurrent exposure to pro-inflammatory states due to repeated surgeries, episodes of sepsis, longer duration of PN, and prolonged periods of fasting. In addition, worsened inflammation in the setting of NEC has been associated with stricture formation (17). In our cohort of patients, in both medical and surgical NEC, cholestasis was associated with an increased rate of bloodstream infections and surgical complications, including stricture formation. Similar to our findings, a retrospective Finnish population study demonstrated that septicemia increased the risk of cholestasis by 60% (RR 1.6: 95%CI 1.0-2.3) in infants with spontaneous intestinal perforation or NEC (9).
Overall compared to patients with medical NEC, infants with surgical NEC had more delay in initiation of feeds. This could have been due to the time taken for return of bowel function after surgery, practice variation that governs when the first feed is initiated, the duration of antibiotics, or general stability of the infants (18). In this setting, some hesitancy and delay in feeds may be related to concern for NEC recurrence with rapid initiation of enteral feeds after recovery from NEC. When analyzing only those surgical NEC patients and comparing those who developed cholestasis versus those who did not, we noted a significant delay in time to first feed and time to reach full feeds in those surgical NEC patients with cholestasis. Determining whether this finding is causal and influences the development of cholestasis or is related to overall patient illness is beyond the scope of this study. A meta-analysis comparing early versus late enteral feeding found that earlier feeding after NEC was associated with a decreased rate of post-NEC complications, including NEC recurrence and/or post-NEC stricture and bloodstream infection (19). However, much of the data presented in that meta-analysis was retrospective in nature, highlighting the need for further understanding regarding the influence of post-NEC feeding and complications, including cholestasis.
The duration of PN had a positive correlation with cholestasis development in both medical and surgical NEC groups. Previous studies have shown a correlation between cholestasis and length of PN therapy. For those patients with surgical NEC, Christensen et al found that 66% of patients who were treated with PN for more than 28 days developed PNAC, increasing to 100% of those patients treated for 57-100 days (6). In our population we defined cholestasis solely based on the degree of direct bilirubin elevation and did not include length of PN therapy. While there was an association with length of PN therapy and the development of cholestasis, we are not able to specify the portion of our population who may have been defined as PNAC versus those who were not. Also, it is important to note our data spanned 2010 – 2016, after which time changes in the prevention and treatment of infants with cholestasis, including newer generation lipid emulsions, would not have been captured. Due to the nature of our database, the details of the components of PN could not be ascertained.
Infants with cholestasis were of larger weight at discharge. We were unable to obtain the z-score for these patients, making interpretation of this finding difficult as again, those patients with cholestasis had longer length of stay and were likely of older postmenstrual age at discharge. In other populations, cholestasis has been associated with poorer growth, with lower weight percentiles at hospital discharge and six-month follow-up (20). Infants with cholestasis in both medical and surgical NEC groups showed a significantly higher prevalence of morbidities related to prematurity including lung disease requiring mechanical ventilation, IVH, PVL, and ROP. In a population-based retrospective case-control study from Sweden the prevalence of bronchopulmonary dysplasia and retinopathy of prematurity in preterm infants with cholestasis were three and eight times more, respectively, compared to those without cholestasis, while rates of IVH were similar between the groups (21).
We observed that an increased peak direct bilirubin correlated with mortality or extreme length of stay. We found that for both groups, peak bilirubin remained a significant predictor of death or eLOS even after accounting for other predictors such as birth gestational age, days of PN, and age at NEC diagnosis. In this setting, the advanced age associated with this outcome could be due to those more premature infants being diagnosed with NEC at an older age, as has been observed by others (15). Others have also reported an increased risk of mortality and direct bilirubin levels (22-24). Overall, in our cohort the rate of mortality for NEC is much lower than has been reported elsewhere in the literature. In a 2020 systematic review, Jones et al. noted an overall mortality rate of 23.5% in all neonates with NEC and a mortality of 34.5% for those that underwent surgery (25). In our analyzed cohort we excluded those infants who died within 14 days after NEC, accounting for a total of 17% (690/4143) of our total population. We excluded these infants as we wanted to study the long-term outcomes of cholestasis in the setting of NEC. This exclusion of early mortality likely accounts for at least a portion of our lower reported overall and surgical mortality rates.
The strengths and limitations of our study arise directly from the use of a multicenter database. One of the major strengths of this study is the use of pooled data from many NICUs, greatly increasing our available sample size. In extracting data from CHND, however, we were unable to establish the timing of peak direct bilirubin in relation to the other comorbidities that could contribute to liver dysfunction and cholestasis. We do know, however, that all included infants had peak direct bilirubin after NEC diagnosis. We were also unable to trend the direct bilirubin levels over time and were limited to a single recorded value. This limited our ability to study the onset and resolution of cholestasis in these infants. Overall, the variables we could choose to assess were limited based on the data collected and recorded in the database. As it is inherent in large databases involving multiple NICUs, we were unable to account for the variability in clinical practices that might affect the variables of interest in our study.
In this study we describe the incidence, risk factors and outcomes of cholestasis in medical and surgical NEC, including predictors of cholestasis development and mortality. Cholestasis is associated with poorer outcomes in both infants with medical and surgical NEC. This important finding suggests that cholestasis in medical NEC is not a benign entity; irrespective of the cause-effect associations, cholestasis in medical NEC deserves the same enhanced monitoring and treatment it receives in surgical NEC.