This study is the largest study to date evaluating associations between PNS and AKI in VLBW neonates. In the current study we show that PNS exposure and AKI occur commonly in this population. Furthermore, we show that in our cohort, PNS are associated with an increased incidence of AKI. However, when evaluated individually, hydrocortisone initially was associated with decreased AKI, but this association was no longer present after multivariable analysis. These results highlight the need for further study evaluating the impact of PNS on the incidence of AKI and associated outcomes.

AKI is a common occurrence in the NICU and associated with increased morbidity and mortality in critically ill neonates. AKI prevalence in infants born extremely prematurely is as high as 48% [1]. These same extremely premature neonates who are at high risk for AKI frequently are exposed to PNS, which could impact kidney health. In the current study we sought to address this knowledge gap by examining associations between PNS and neonatal AKI using a large cohort of VLBW neonates cared for in a NICU with protocolized AKI care. Our findings add to the findings of prior studies of neonatal AKI [22] and corroborate those of other investigators. We found neonatal AKI is significantly more common in infants with lower GA, BW, and APGAR scores, those with AKI had significantly longer LOS and those with AKI were significantly more likely to die prior to discharge. Independent risk factors for AKI in our cohort included hypotension, PDA, sepsis, and mechanical ventilation. Similarly, we found subjects treated with PNS had significantly lower GA, BW, and Apgar scores at 1 and 5 min, had significantly longer LOS, and were significantly more likely to die prior to discharge. Our novel findings include a 1.72-fold increased adjusted relative risk of developing AKI in those infants exposed to PNS, which was contrary to our hypothesis. We also found that infants with AKI were significantly more likely to have been diagnosed with AI than those without AKI, though we note that AI could certainly have gone undiagnosed in our cohort, particularly in those to whom PNS were prescribed. Lastly, in performing a sub-analysis on HC exposure and associations with AKI, we detected a protective association between HC exposure and AKI; however, after controlling for significant confounders, this association was no longer present.

Our hypothesis that PNS exposure is associated with decreased incidence of AKI was based on both the findings of Baserga et al. and Shimokaze et al., as well as the underlying physiology of relative or transient AI [11, 12]. Shimokaze et al. investigated late-onset glucocorticoid-responsive circulatory collapse (LCC) in VLBW infants, an event characterized by sudden onset hypotension and/or oliguria after 7 days of life, which is resistant to volume expanders and inotropes but responds rapidly to intravenous glucocorticoids. They found that 14 of 41 infants (34%) in their cohort developed LCC with AKI (primarily documented with decreased UOP) and all of these infants had resolution with the initiation of HC. Baserga et al. looked similarly at VLBW infants during this time frame that developed AKI, however without the circulatory collapse. They found that 27 of 230 (11.7%) infants developed AKI without clear underlying etiology, all of whose AKI resolved with the initiation of HC. Both groups concluded that relative AI in these infants could be the cause of their AKI. Further articles describe LCC and investigate the pathophysiology, hypothesizing again its relation to relative adrenal insufficiency in this population [13, 14]. Fofi et al. described the importance of the renin-angiotensin-aldosterone system (RAAS) in renal homeostasis and the role that RAAS disruption may play in the development of AKI in adult patients. In patients with AI, a combination of low aldosterone levels causing hypovolemia and low cortisol levels leading to decreased tone of the efferent renal arterioles with subsequent reduced perfusion pressure could together lead to AKI [23]. Although not reported, logic suggests that this mechanism would apply to preterm infants as well. Extremely preterm neonates likely also have an immature hypothalamic-pituitary-adrenal (HPA) axis. The HPA axis is functional in the fetus by 20 weeks of gestation but does not fully mature until after 30 weeks of gestation. The administration of exogenous glucocorticoids suppresses cortisol production and secretion of cortisol through negative feedback on this immature HPA axis, potentially leading to suppression of this HPA axis and subsequent AI [24].

Our findings did not support our hypothesis. Rather than being protective against AKI, we found that PNS exposure is associated with increased risk of AKI development. We suspect the detected association between PNS exposure and AKI may be partially confounded by indication and the inclusion of multiple types of PNS. Despite our efforts to control for degree of illness and AKI confounders in our analysis, it is possible that PNS exposure is a marker of clinical severity of illness in these patients and subsequently their increased risk for AKI. However, the median length of time between steroid initiation and AKI diagnosis of 12 days indicates a prolonged separation between the presumed acute illness necessitating start of steroids and an AKI diagnosis, potentially making this confounding less likely. Given the potential links between the RAAS pathway, the immature HPA axis and AKI in the premature infant, as discussed above, it is possible that in some cases, neonates may experience underlying undiagnosed AI; when these infants receive PNS, it may serve a renal protective effect while the infant is exposed, and AKI occurs in these infants when the steroids are discontinued leading us to detect a misleading association between PNS exposure and AKI. The prolonged median 12-day period between steroid initiation and AKI diagnosis may additionally lend merit to this hypothesis, as possibly these infants had begun a steroid wean or even had their steroids discontinued during this timeframe. AI is difficult to diagnose in the premature neonate, as cortisol levels do not appear to correlate with disease severity and most infants are diagnosed based on clinical symptoms (hyponatremia, hyperkalemia, hypotension, oliguria) [25]. This may lead to an underdiagnosis of AI which may cause an under recognition of infants at risk for development of AKI post-PNS exposure. Though we found a strong association between AI and risk of AKI in bivariable analysis, this association was no longer present in adjusted analysis potentially due to underdiagnosis of AI.

In a sub-analysis of associations between solely HC exposure and AKI, a protective association was found between HC exposure and AKI. However, after controlling for significant confounders, HC exposure alone was no longer associated with AKI. There are several potential reasons why we found an association with all PNS exposure, but not HC alone. First, dexamethasone, the most common steroid in the PNS exposure group besides hydrocortisone, is 40-50 times stronger than HC and longer-acting [21], potentially creating more suppression on the HPA axis and leading to an iatrogenic AI in these infants. Second, as only 68 of 567 infants received HC, there is a potential the sample size was too small to detect an association, and with a larger sample size we may have seen the protective association between HC and AKI stand even in an adjusted analysis.

Limitations of our study include the observational and retrospective design. Intervals between serum creatinine evaluations were at provider discretion and not protocolized. Similarly, PNS prescription was at the discretion of the provider. In the absence of a more detailed understanding of the timing, indication for, and type of steroid administered, conclusions based on the detected association between steroid exposure and AKI are limited. However, given the scarcity of available literature and studies examining this association, these findings bear consideration. Though we determined that several confounding factors were associated with risk of AKI, additional confounders may exist that were not examined. Additionally, AI diagnosis was also at the discretion of the provider and can be difficult to diagnosis.

In conclusion, the current study shows that PNS exposure was associated with a 1.72-fold increased adjusted relative risk of developing AKI in our cohort. HC exposure alone was not associated with increased risk of AKI. These findings should be interpreted with caution and warrant further detailed study as the relationship between AI, PNS, and indication for PNS treatment is extraordinarily complex. In future work, it will be important to investigate both timing, type, and medical indication for PNS exposure when examining associations with AKI, as well as including timing of AKI whether it be closely following delivery (as in late onset glucocorticoid-responsive circulatory collapse) or later in the infant’s course. It will be important to examine PNS dosing and duration of treatment, as well as temporal association with AKI diagnosis including if the patient was still on PNS at time of AKI and/or length of time from discontinuation to diagnosis of AKI. It may also be prudent to examine re-initiation of PNS and subsequent resolution of AKI. It will also be important to utilize consistent criteria to diagnose AI to evaluate associations more effectively.