Introduction

Preterm birth remains a significant problem in the USA. In 2014, over 55,000 preterm infants were born with a very low birth weight (VLBW <1,500 g) [1]. Advances in technology, medications and neonatal intensive care unit (NICU) treatments have decreased the mortality rate in these infants, but the co-morbidities of preterm birth remain common [2–4].

Chronic kidney disease (CKD) has been shown to occur more frequently in LBW infants (LBW <2,500 g) [5]. LBW infants are often born preterm and are at risk of beginning life with a nephron deficit, as typically 60% of nephron development occurs in the third trimester [6]. An autopsy study of preterm infants demonstrated ongoing nephrogenesis following birth limited to 40 days, which represents early cessation of nephrogenesis in most of these infants [7]. Signs of kidney dysfunction in former preterm infants can be detected in childhood and presents as decreased renal volume, elevated blood pressure (BP) or microalbuminuria [8–12]. However, the only screening recommendation for renal follow-up is by the American Academy of Pediatrics (AAP), advocating BP measurement prior to 3 years of age in VLBW infants [13].

Although the renal consequences of preterm birth are becoming apparent, less is known regarding the long-term effects of acute kidney injury (AKI) in this population. AKI in VLBW infants is common and ranges from 12 to 40% depending upon the definition used and the composition of the population studied [14–17]. AKI was previously thought to be a reversible insult; however, many studies over the past decade have demonstrated that AKI results in an increased risk of developing permanent damage and ultimately CKD in animal models and humans [18–26]. A recent publication summarized eight longitudinal studies that evaluated the long-term renal function of neonates exposed to AKI [27]. Of the 293 children assessed, 53 (18%) had evidence of CKD with an estimated glomerular filtration rate (eGFR) <90 mL/min/1.73 m2 at follow-up. However, these observational studies were primarily comprised of full-term neonates with limited representation of the VLBW population. The goal of the FANCY study was to evaluate the effect of AKI on long-term renal function in a previously characterized preterm and VLBW cohort.

Materials and methods

Study design

A prospective cohort study was conducted from 1 November 2014 to 1 March 2016 at the University of Virginia (UVa) and approved by the Institutional Review Board. A guardian of each participant provided written informed consent for one research visit.

Participants

Subject recruitment originated from a retrospective study of preterm VLBW subjects born between 2008 and 2011 [16]. The parents of potential subjects were contacted by phone and their children recruited. Inclusion criteria were birth weight ≤1,500 g and UVa NICU admission before 2 days of life. Patients with congenital anomalies of the kidney or urinary tract were excluded.

Data collection

Maternal and neonatal factors from the NICU hospitalization were collected from the electronic medical record (see Table 2). Maternal factors included betamethasone exposure and the presence of preeclampsia or diabetes. Neonatal factors included a clinical risk index for babies (CRIB II) score, intraventricular hemorrhage (IVH, grade III or IV), bronchopulmonary dysplasia (BPD, Vermont Oxford Network (VON) definition [28]), steroid exposure (hydrocortisone or dexamethasone), medical or surgical treatment of patent ductus arteriosus (PDA), necrotizing enterocolitis (NEC, modified Bell’s stage 2 or greater), and culture positive sepsis (blood or urine culture). Medications received in the NICU were recorded as either total number of milligrams per kilogram for the hospitalization or number of days (see Table 2). Kidney-specific information collected from the NICU included the number of AKI episodes, highest stage of AKI, peak creatinine, number of days in which serum creatinine was >1 mg/dL, initial and last creatinine before discharge. AKI was classified using the modified Kidney Disease Improving Global Outcomes (KDIGO) definition excluding urine output (Table 1) [29].

Table 1 Kidney Disease: Improving Global Outcomes (KDIGO) acute kidney injury (AKI) definition modified to exclude urine output
Full size table

During the study visit height (cm) and weight (kg) were measured and body mass index (BMI) was calculated. Blood pressure was measured in a seated position in the right upper extremity with an appropriate cuff [13]. A trained sonographer obtained the length, width, and anterior–posterior diameter of each kidney using a Philips EPIQ 7G. Kidney volume was calculated using the ellipsoid formula 1/6*π*kidney length(cm)*kidney width(cm)*kidney height(cm) [30].

Serum creatinine was measured at the UVa clinical laboratory using the alkaline picrate (Jaffe) method traceable to isotope dilution mass spectrometry. Cystatin C was determined by Mayo Medical Laboratories using the internationally standardized particle-enhanced turbidimetric assay (PETIA) [31]. Urine was obtained by clean catch for measurement of protein (Upro assay using benzethonium chloride as the protein denaturing agent) and creatinine. The urine protein/creatinine (UPC) was calculated. eGFR was calculated by the original Schwartz equation [32] and by GFR = 77.24 × (cystatin C mg/L)−1.2623 [31, 33]. Renal dysfunction was defined as the presence of any of the following: eGFR <90 mL/min/1.73 m2, UPC >0.2 or BP >95th %tile.

Study size

Of the 455 subjects previously studied from 2008 to 2011 [16], 396 were alive (mortality 13%); 154 had a history of AKI and 242 had no AKI. Given the lack of previously published data on the development of CKD in former VLBW infants following KDIGO-defined AKI in the NICU, a power analysis could not be performed from the literature. Therefore, a power analysis was performed using the first 8 subjects (5 AKI, 3 no AKI). Given the standard deviations from preliminary data, with 30 (19 with AKI, 11 without AKI) subjects, the Mann–Whitney U test would have 80% power with a two-sided level of significance of 5% to detect a 23 mL/min/1.73 m2 difference in eGFR between groups.

Comparison with term cohort

To determine the effect of preterm birth on renal function, we compared our cohort’s cystatin C values with those in a previously published cohort of term infants who were studied at 6 years of age using the PETIA method [34].

Statistical analyses

Mann–Whitney U or t tests were used to compare continuous variables and Chi-squared or Fisher’s exact tests were used to compare categorical variables. Logistic regression was used to estimate the association between AKI and the development of renal dysfunction. Individual models were built using 36 potential confounding variables.

Results

Study cohort

Parents of 167 previously studied children were contacted between November 2014 and March 2016. Forty-two parents consented to participate and 34 children completed all elements of the study visit (Fig. 1).

Fig. 1
figure 1

Subject recruitment flowsheet. This diagram outlines the available study population recruited to the Follow-up of Acute Kidney Injury in Neonates during Childhood Years (FANCY) study and those that completed the study. 1 n = 23 not called, n = 3 died, n = 4 no phone number, n = 36 out of area, n = 2 exclusion criteria. 2 n = 62 not called, n = 7 died, n = 28 no phone number, n = 52 out of area, n = 12 exclusion criteria

Full size image

Clinical demographics from NICU hospitalization

Of the 34 subjects who completed the study, 20 had a history of AKI in the NICU (stage 1, n = 8; stage 2, n = 9; and stage 3, n = 3) and 14 did not have AKI. Complete clinical demographic factors are shown in Table 2. Maternal preeclampsia was more common in the group without AKI. The AKI group had lower gestational ages, lower birth weights, higher CRIB II scores, longer hospitalizations, more ventilator days, more treated PDAs, and more culture-positive sepsis. Subjects with AKI also received more indomethacin, gentamicin, vancomycin, and caffeine than subjects who did not have AKI. Subjects in the AKI group had a higher median peak creatinine during their NICU hospitalization, more days with a creatinine level above 1.0 mg/dL, and lower initial creatinine.

Table 2 Neonatal intensive care unit (NICU) hospitalization demographics
Full size table

Clinical data at follow-up evaluation

Clinical data at the time of follow-up are summarized in Table 3. The median age at follow-up for each group did not differ. The mean z-score for height was lower in the AKI group than in the group without AKI.

Table 3 Kidney health evaluation results
Full size table

There were no differences in serum creatinine or cystatin C between the AKI group and the group without AKI. Similarly, there was no difference in eGFR using either creatinine or cystatin C (Fig. 2). The median eGFR calculated by cystatin C for the entire cohort was lower than the median eGFR calculated from creatinine using the Schwartz equation. There were 9 subjects in the entire cohort with an eGFR (cystatin C) <90 mL/min/1.73 m2 and no subjects with an eGFR <90 mL/min/1.73 m2 using serum creatinine. When eGFR was analyzed using cystatin C and stratified by stage of AKI, there were no statistical differences between stage of AKI and eGFR (Fig. 3). In addition to the 9 subjects with an eGFR <90 mL/min/1.73 m2, 5 subjects in the entire cohort demonstrated hyperfiltration, defined as >120 mL/min/1.73 m2 (AKI, n = 4; no AKI: n = 1, p = 0.38).

Fig. 2
figure 2

Comparison of renal function using serum creatinine and cystatin C. The two graphs on the left compare creatinine at the time of follow-up and estimated glomerular filtration rate (eGFR) at the time of follow-up as calculated by creatinine. The groups are divided into those who experienced acute kidney injury (AKI) in the neonatal intensive care unit compared with those who did not. Individual values are displayed in addition to median and IQR for the groups. The graphs on the right compare cystatin C and eGFR as calculated by cystatin C at the time of follow-up. The dotted line on the two lower graphs displays 90 mL/min/1.73 m2, stage 2 chronic kidney disease

Full size image
Fig. 3
figure 3

Function according to AKI stage. eGFR was calculated using cystatin C and each individual measurement plotted by peak stage of AKI while in the neonatal intensive care unit. Median values for each stage are displayed by the horizontal line

Full size image

Systolic and diastolic blood pressure measurements did not differ between the groups and, as shown in Table 3, subjects in the AKI group had significantly higher UPC. There were no differences in the raw or the adjusted renal volumes [for height or body surface area (BSA)] between the two groups. One subject in the AKI group had increased echogenicity in the upper pole of the left kidney.

Comparison with term cohort

When the preterm AKI and no AKI cohorts were compared with a term group of infants in whom cystatin C was measured using the same PETIA methods, the cystatin C values were significantly different (0.84 mg/L, 0.80 mg/L, 0.78 mg/L respectively, p = 0.006) [34]. The combined preterm cohort (AKI and no AKI groups) also had a higher mean cystatin C value than the term cohort (0.82 mg/L vs 0.78 mg/L, p = 0.007).

Renal dysfunction (primary outcome)

In the AKI group, there were 4 subjects with an eGFR <90 mL/min/1.73 m2, 4 with UPC >0.2, 2 with BP >95th percentile, and 1 subject with both decreased renal function and elevated BP. The relative risks for eGFR <90 mL/min/1.73 m2, UPC >0.2, and BP >95th percentile were 1.5 (95% CI 0.8–2.5), 1.9 (95% CI 0.9–2.6), and 1.8 (95% CI 0.8–10) respectively. Sixty-five percent of the AKI group had renal dysfunction compared with 14% in the group without AKI (p = 0.01). The relative risk for developing renal dysfunction was 4.5 (95% CI: 1.2–17.1, p = 0.01) times higher in the AKI group.

There were no significant differences in z-scores for weight or height between subjects with and those without renal dysfunction (Table 4). As outlined in Table 4, those with renal dysfunction received more indomethacin and caffeine, had a higher peak stage of AKI, more episodes of AKI, higher peak creatinine, and more median days with creatinine above 1 mg/dL.

Table 4 Comparison of subjects with and those without renal dysfunction
Full size table

Univariate confounder analysis

We explored the association between current renal dysfunction and neonatal AKI by adjusting individually for each of 36 possible confounding variables (see Table 2 for the list of confounders). Ideally, several or all of these factors would have been accounted for simultaneously, this was not possible within the limits of our sample size. Each factor, with the exception of neonatal sepsis, attenuated the effect of AKI for predicting renal dysfunction. However, only gestational age, CRIB II score, and treated PDA eliminated the association between AKI and future renal dysfunction. After adjustment for the 33 other factors, AKI remained a significant risk factor for developing renal dysfunction at the age of 5 years.

Discussion

In this prospective cohort follow-up study of AKI in preterm VLBW infants, 65% of children with a history of AKI had renal dysfunction at a median age of 5 years, whereas 14% of children with no history of AKI had renal dysfunction (p = 0.01). The subjects with renal dysfunction were more likely to have had a higher stage of AKI, more episodes of AKI, a higher peak creatinine, and more days with a serum creatinine above 1 mg/dL while in the NICU. The conventional biomarker of renal dysfunction, serum creatinine, failed to detect any subjects with an eGFR <90 mL/min/1.73 m2, whereas the use of the biomarker cystatin C demonstrated that 26% of the study population had an eGFR <90 mL/min/1.73 m2.

There are few studies that assess the long-term renal health of preterm neonates. The available literature is comprises of single center studies with heterogeneous patient populations and practice patterns. When AKI is assessed, there is significant practice variation in monitoring serum creatinine in the NICU, multiple definitions of AKI, and variable outcome assessment methodology. These are all key factors in the comparison and generalizability of this study to others in the literature. In the current study, 26% of the VLBW cohort (AKI and no AKI groups) had an eGFR <90 mL/min/1.73 m2. This is comparable to the study by Nishizaki et al., which reported in 2014 that 38% of children born VLBW had CKD (19 out of 50), with an eGFR <90 mL/min/1.73 m2 [35]. Although Nishizaki et al. did not stratify their VLBW group by AKI history, their median cystatin C (0.85 mg/L) was similar to that of the FANCY AKI group (0.84 mg/L). The first follow-up study in preterm children where the impact of AKI was acknowledged was published in 2003 [36]. Abitbol et al. described a cohort of 20 extremely LBW infants with a significant AKI history (defined by a serum creatinine >2.0 mg/dL for >48 h and/or oliguria defined by UOP <0.5 mL/kg/h for >24 h) [36]. The results of our studies cannot be directly compared given the different definitions of AKI, the significant differences in mean peak creatinine for the cohorts (3.2 mg/dL vs 1.2 mg/dL), and the difference in age at the time of follow-up (9 years vs 5 years). However, it is significant to note that CKD is common; 45% of this population went on to develop CKD with an eGFR <90 mL/min/1.73 m2 at the average age of 9 years (eGFR: 29 ± 22 mL/min/1.73 m2).

Recently, Bruel et al. published follow-up results of preterm patients with a history of AKI using creatinine cut-offs for gestational age groups (GA 24–27: 1.6 mg/dL, GA 28–29: 1.1 mg/dL, GA 30–32: 1.0 mg/dL) [37]. The mean follow-up age was 6.6 years and there were 25 children with a history of AKI and 49 children without AKI. Similar to our study, there were no differences in eGFR between the groups when serum creatinine was used to evaluate function, but despite the use of serum creatinine, nearly a quarter of their population (23%) had an eGFR <90 mL/min/1.73 m2, regardless of AKI status. In contrast to the present study, the Bruel group demonstrated no difference in protein excretion between the AKI group and the group without AKI, but did detect smaller renal volumes in their AKI group. Although there are some similarities between the two studies, there are significant differences in the AKI definition used and outcome metrics. Larger studies are necessary to determine the most precise definition of neonatal AKI. Future studies require rigorous controls to decipher the effects of AKI, preterm birth, and subsequent therapies on long-term renal function. In combination, these studies provide accumulating evidence that VLBW neonates are at an increased risk of renal dysfunction during childhood and help to inform a more precise definition of AKI in the neonatal period, which is based on long-term renal outcomes.

Our study, FANCY, also unmasks additional novel findings. The FANCY study included neonates with stage 1 AKI, defined by small changes in serum creatinine. The AKI group contained 8 children with stage 1 AKI; 5 of whom had evidence for renal dysfunction and 2 had an eGFR <90 mL/min/1.73 m2. The adult AKI literature supports the notion that even small changes in serum creatinine confer a higher risk for mortality [38], but it is still the opinion of many neonatologists that stage 1 AKI may not confer the same risk for mortality or development of CKD as stages 2 or 3 [39]. The FANCY study was not sufficiently powered to determine if stage 1 AKI is an independent risk factor for the development of CKD, but the data suggest that neonates with stage 1 AKI should be included in larger multicenter prospective trials to determine the specific long-term risk of mild AKI.

Another interesting observation in the FANCY cohort, and highlighted in previous AKI studies, was that VLBW neonates born to mothers with preeclampsia are seemingly “protected” from AKI [15, 40]. There are several hypotheses as to why these neonates might be protected from AKI. First, it has been postulated that neonates born to mothers with preeclampsia may be healthier, as birth was induced for maternal reasons. Second, magnesium therapy may be directly protective [17]. In our study, 3 neonates in the AKI group (15%) and 8 neonates in the no AKI group (57%) were born to mothers with preeclampsia (p = 0.02). At the time of follow-up, of the three neonates in the AKI group, 1 developed proteinuria, and their eGFRs by cystatin C were 92, 123, and 92 mL/min/1.73 m2. In the group without AKI, 2 of the 8 neonates born to mothers with preeclampsia had eGFRs below 90 mL/min/1.73 m2. The numbers of subjects in this study is too low to determine if neonates born to preeclamptic mothers suffer long-term renal damage, despite a lower incidence of AKI, but this deserves future attention in prospective studies.

We have demonstrated a significant height difference between the AKI group and the group without AKI at the time of follow-up. The height Z-score in the AKI group was lower than in the group without AKI. The difference in height is particularly pertinent, as current estimates of GFR include height. In the FANCY cohort, 5 subjects in the AKI group and 1 subject in the no AKI group had negative z-scores. Although it is not possible to determine causation, AKI is associated with a hypercatabolic state and therefore may directly contribute to the development of poor linear growth. Future work should include detailed nutritional assessments and fluid balance metrics to more fully elucidate the relationship between neonatal AKI and growth.

A secondary objective of the FANCY study was to examine neonatal risk factors that may help in the identification of patients who may need long-term renal follow-up. Serum creatinine days >1.0 mg/dl, stage of AKI, the number of AKI episodes, and the peak serum creatinine were all associated with the development of renal dysfunction. Previous studies have suggested that the increase in creatinine seen in VLBW infants following birth is normal, owing to transient reabsorption across immature tubules [41, 42]. However, Weintraub et al. challenged this tenet and theorized that the early rise in creatinine may indeed be AKI and not a normal physiological response [17]. Our data further support this finding, given that our subjects with renal dysfunction at follow-up had a modest median peak creatinine of 1.2 mg/dL—a value that many may consider part of the normal physiological peak. Furthermore, the group without renal dysfunction had far fewer days with a creatinine above 1.0 mg/dL, supporting the Weintraub theory that what was previously thought of as a normal physiological rise in creatinine may indeed be AKI as it is associated with long-term changes. Our study suggests that this rise and fall of creatinine in the first week of life in extremely preterm neonates may not be classified as AKI by current definitions, but may ultimately be an important predictor of long-term outcomes. Although it may be common for VLBW infants to have an early rise in creatinine above 1.0 mg/dL, our study suggests that if creatinine stays above 1.0 mg/dL, the risk for future renal dysfunction may be higher.

In general, there is poor long-term renal follow-up of neonatal AKI, despite KDIGO guidelines that recommend patients with a history of AKI have a CKD evaluation 3 months after their AKI event, and AAP recommendations that VLBW patients have their BP checked at visits up to the age of 3 years [13, 43]. There are challenges in translating these general guidelines to neonates who have experienced AKI. Our findings support the KDIGO recommendations that neonates with a history of AKI merit follow-up, but the optimal age of this follow-up remains unknown. As 65% of the VLBW population with AKI had abnormalities at the age of 5 years, we believe that follow up should occur prior to this age. We agree with the AAP recommendation of BP checks in former preterm children up to the age of 3 years, as elevated BP can be a presenting sign of early CKD. However, BP measurement should not be the only screening method. In the FANCY study, only 3 of the 34 patients had elevated BP, whereas many more patients had either reduced renal function or proteinuria with a normal BP.

Our study has several limitations. First, this is a single-center study with a small sample size and our findings require validation by a multi-center study. Although the number of subjects was small, this NICU population has been extensively studied in a NICU environment where serum creatinine was measured frequently. Second, although we evaluated a narrow age range, there may have been a residual component of time-dependent or age-related changes in outcomes. Third, our study lacked the gold standard measurement of GFR (e.g., iohexol), and the cystatin C equation used to estimate GFR was not based on pediatric subjects, but is the recommended equation to be used given the method of testing and validation by Mayo Labs. Although there may be differences in comparison with other studies, this would not affect our comparison of absolute differences in cystatin C values. Our study lacked a full-term control group to compare renal function according to cystatin C. From the literature, we found a large control group where children born at full term had an assessment of renal function by cystatin C using the same PETIA methodology [34]. This full-term control group had a mean cystatin C of 0.78 mg/L, significantly lower than the value for our group without AKI (0.80 mg/L), suggesting that there might be small differences between term and preterm children without AKI exposure when large populations are compared. Finally, although 42% of the surviving 396 infants were contacted, only 10.5% of the children participated in the follow-up visit, limiting the generalizability of the results.

In conclusion, in our single-center study, neonatal AKI in VLBW infants increased the risk of renal dysfunction in childhood. Our data strengthen the concept that cystatin C is a more sensitive biomarker for the detection of subtle renal dysfunction, and may be an important biomarker in the former preterm infant population to detect early signs of renal disease. Small changes in serum creatinine that comprise the KDIGO stage 1 definition of neonatal AKI may be important in the long-term renal outcomes of VLBW neonates [23]. It is clear that further work in this area is needed, but it is important to remember that the VLBW group represents a large patient population and often appears quite healthy as they distance themselves from their challenging preterm birth. Pediatricians and pediatric nephrologists should be aware that AKI is a risk factor for CKD and children with a history of VLBW and AKI may require more vigilant monitoring [27]. If modifiable abnormalities such as hypertension, obesity, smoking or proteinuria are identified in patients at risk for CKD, lifestyle modifications such as exercise, weight loss, and salt restriction, in addition to medications to treat hypertension and proteinuria can be recommended. These findings should motivate further large prospective studies to evaluate potential therapies to prevent and treat AKI in the VLBW population so that kidney health and function in this vulnerable population can be preserved.