Characterizing dynamics of serum creatinine and creatinine clearance in extremely low birth weight neonates during the first 6 weeks of life

The availability of reference ranges for any specific laboratory test or biomarker to support clinical decision-making and tailor therapy to the individual patient will greatly support neonatal care [1, 2]. Variability is a key feature in the neonatal population since maturational physiological changes are most prominent during early infancy. This variability is related not only to differences in current weight, gestational age (GA), or postnatal age (PNA), but also to morbidities, co-medication, or nutritional and fluid management. Throughout the neonatal period, serum creatinine (Scr) concentrations in preterm neonates vary tremendously due to this large inter- and intra-individual variability [3, 4].

Despite its limitations, Scr is a commonly measured and easily accessible biomarker to estimate glomerular filtration rate (GFR). The clearance of Scr defines the volume of blood plasma that is cleared of creatinine per unit time and approximates the GFR. Many mathematical equations have been proposed to reflect creatinine clearance or estimated GFR in newborn infants, mostly by means of linear regression methods [4‐8]. In classical linear regression, there is only one level of unexplained variability (difference between observation and predicted value). In contrast, applying modeling and simulation techniques rather than classical linear regression methods allows describing and quantifying the maturation processes underlying neonatal physiology with multiple levels of variability. These population models are characterized by the typical individual (the mean) and the random effects, describing the variability of the data [9]. These random effects are divided into two levels: the difference between individuals (inter-individual variability, IIV) and the difference between the individual prediction and the observation (individual prediction error, comprising also analytical imprecision of the observation).

Although the developed equations have supported clinicians in the assessment of estimated GFR, the physiology and normative range of Scr in extreme preterm infants are still not fully understood and the accurate assessment of kidney function remains challenging. Until now, we are not aware of any mathematical modeling effort applying modeling and simulation techniques to Scr data in neonates.

Traditionally, the dogma has been to ignore elevated Scr in preterm infants during the first days after birth as they are considered uninformative about creatinine clearance due to maternal Scr transfer. Additionally, passive reabsorption across immature leaky kidney tubules contributes to transient accumulation of creatinine during the first days of life, which suggests that part of elevated Scr after birth may still reflect neonatal and not maternal kidney function [10].

Determining and establishing (gestational) age-dependent Scr reference ranges early in neonatal life could support the clinical evaluation of kidney function, especially when considering extremely low birth weight neonates (ELBW; ≤ 1000 g). The aims of this study were (i) to characterize the dynamics of Scr and associated creatinine clearance during the first 6 weeks of life by the development of a mathematical model, (ii) to identify and quantify factors influencing Scr and kidney function in this specific population, and (iii) to derive age-dependent Scr reference ranges to facilitate monitoring and evaluation of kidney function in ELBW neonates.

Data from 158 ELBW neonates were collected, of whom ten qualified as dropouts (no information on GA or MOD) and were excluded from the analysis. Over a period of 6 weeks after birth, a total of 2814 Scr concentrations of 148 ELBW neonates were included in the model development analysis. Sixty-nine ELBW neonates were included in the external evaluation analysis with 1212 Scr concentrations.

This is the first study that provides reference ranges for real-world Scr concentrations and associated creatinine clearance for ELBW neonates by characterizing the dynamics of Scr concentrations during the first 6 weeks of life in a quantitative manner by means of a mathematical model. GA, MOD, and ibuprofen treatment were key factors influencing individual creatinine profiles and corresponding creatinine clearance, while prenatal lung maturation and the need for treatment with inotropic agents were not significant covariates. There are several aspects that require further interpretation and context.

First, in 2002, Léger et al. have applied population pharmacokinetic approaches to pediatric creatinine data with the aim of characterizing GFR. Their population comprised children with a mean age of 11 years, making the results not translatable to our preterm neonatal population [7]. Vieux et al. obtained GFR reference values retrieved from regression models based on Scr assessments in plasma and urine on day 7, 14, 21, and 28 of life [4]. Their assessment of median GFR for a 27-week-old neonate on day 14 after birth was higher as compared with our calculations (16.22 ml/min/1.73 m² vs. 9.40 ml/min/1.73 m²; Table 4). This difference might be explained by the fact that our population was considered more immature (smaller GA and lower birthweight), and different equations were applied for the calculation of individual BSA (Dubois and Dubois versus Ahn) [16, 18]. In the current study, BSA was calculated based on current weight measurements, reflecting an accurate image of postnatal development. Furthermore, it is well known that bilirubin is one of the most relevant interfering compounds when creatinine is measured. As bilirubin is mainly an issue in early neonatal life, and subsequently disappears, this also indicates that creatinine decrease in absolute values will be more pronounced over the first week(s) of neonatal life, suggested to reflect a higher creatinine clearance. Additionally, different assays were used for the Scr quantification (Jaffe versus enzymatic) [4, 11]. It has been previously shown that Jaffe Scr concentrations were significantly and systematically higher (difference 0.1–0.2 mg/dl) compared with the enzymatic quantification throughout postnatal life [19]. Along the same line, there is a difference of about 0.1 mg/dl in urinary creatinine measurements between a Jaffe and an enzymatic assay [20]. The advantage of our study is that we characterized the dynamics of Scr concentrations over the entire 6 weeks after birth in a quantitative manner rather than comparing the GFR values on separate days in ELBW cases.

Second, controversy remains regarding the first creatinine concentration at birth. We observed lower Crea_birth concentrations in the more extreme preterm neonates, who are born during the second trimester (median (95% CI); 0.50 mg/dl (0.31–0.81) at 24 weeks GA; Table 4). ELBW neonates who are born in the third trimester (> 28 weeks) were characterized by higher Crea_birth Scr (median (95% CI); 0.79 mg/dl (0.50–1.29) at 32 weeks GA). This trend is also, albeit to a lesser extent, reflected by the maternal creatinine concentrations, which are generally, but marginally lower during the second trimester (0.59 mg/dl) as compared with the third trimester (0.61 mg/dl) [21].

Third, we illustrated a clear relationship between creatinine clearance and Scr concentrations, starting after the first postnatal days (Fig. 2). Adaptation to extra-uterine life or maternal Scr contribution might be potential hypotheses for lack of relationship between creatinine clearance and Scr concentrations on the first day after birth, although this requires further investigation. Overall, it demonstrates that Scr concentrations, which are routinely collected directly after and during the first days after birth, can be interpreted and can provide crucial understanding on the developing kidney function. In addition, we recognize the unit policy was to obtain Scr concentrations frequently. This can be illustrated by comparing our median number of observations (n = 20/patient) to a recent reference study, like the Assessment of Worldwide Acute Kidney Epidemiology in Neonates (AWAKEN) study, where the median number of Scr concentrations was ≤ 3 and ≤ 5 per patient, respectively in 10 and 15 of the contributing units [22].

Fourth, in this study, antibiotic treatment (e.g. aminoglycosides) has not been assessed as a covariate on any of the model parameters. Since most ELBW neonates in the NICU receive this type of empiric treatment in early neonatal life, we considered this as a nondistinctive feature of our population. The same holds true for nutritional and fluid management strategies. As these interventions are standardized, we were unable to explore these factors. Continuing with drug effects, this study confirms that ibuprofen treatment is associated with a transient decrease in creatinine clearance, although this effect is relatively small (Fig. 4). In contrast, the prescription of inotropic agents was not associated with changes in Scr or creatinine clearance in these datasets. It has to be acknowledged that the dose or exact ibuprofen administration time point during the day is not taken into account; ibuprofen treatment was only assessed as a binary factor on the creatinine clearance for each postnatal day. As such, our analysis may underestimate the effect of ibuprofen treatment on kidney function. Utilized ibuprofen dosing regimen was according to label (10 mg/kg, followed by 5 mg/kg/day every 24 h, for two to four additional days) and was not adapted for the proven increase in ibuprofen clearance with increasing postnatal age [23]. However, we confirm that the model as constructed is indeed fitted to explore potential differences in the incidence and severity of kidney impairment and/or acute kidney injury (AKI) between different dosing strategies.

Fifth, the effect of MOD on Scr concentrations and associated creatinine clearance in our analysis has not been shown before, but co-aligns with the AWAKEN study data, as these authors also observed a positive effect of C-section on the incidence of AKI [22, 24]. It is known that MOD is one of many factors that can influence neonatal physiology, as the recent study by Charlton et al. showed that neonates who were born by scheduled C-section were subjected to a 30% lower probability of acute kidney injury [24]. As acute kidney injury was defined as a Scr increase (0.3 mg/dl), this finding is in accordance with our observations as higher Scr concentrations and lower creatinine clearances were observed for ELBW neonates who were born by vaginal delivery (Figure S2). It has been shown that infants delivered by C-section are subjected to a higher risk for pulmonary hypertension due to the blunted decrease of pulmonary vascular resistance compared with vaginal delivery [25]. Therefore, it might be probable that MOD modulates kidney (function) during early neonatal life since it depends on the vasoconstriction of renal vasculature and its postnatal adaptation pattern, although this hypothesis requires further investigation and more controlled studies. Besides neonatal physiology, maternal morbidities such as preeclampsia and hypertensive disorders are often underlying causes for performing C-section. Whether these maternal morbidities are the cause of lower creatinine concentrations during postnatal life needs to be investigated. However, it is reasonable to anticipate that such maternal morbidities would rather result in higher neonatal Scr concentrations in early neonatal life due to the higher maternal Scr associated with these morbidities [3].

Sixth, our study showed that increasing GA is associated with a steeper increase in GFR over the postnatal period (shorter t₅₀). No inter-individual variability could be estimated on kin_production as it was assumed that the production rate of creatinine remained constant for our ELBW population due to the fact that creatinine is a product that is generated exclusively from muscle metabolism. Previous studies have shown that the fraction of muscle mass does not differ substantially across gestational ages within ELBW cases [26, 27], suggesting that a weight-normalized kin_production could have been appropriate. Our estimate of 3.55 mg/day for an average 0.7 L/kg ELBW neonate corresponds to a weight-normalized production rate of 5.1 mg/kg/day, which is approximately half of adult values and can be considered consistent since fractional muscle mass is approximately half that of adults. Additionally, immature kidney tubules in ELBW neonates have poor concentrating ability [28]. As such, these immature and leaky kidney tubules are partially responsible for passive tubular reabsorption of the filtered and accumulated Scr, mostly during the first postnatal days (Figure S3) [10, 29]. The ratio of the amount of creatinine produced versus eliminated over time conceptually illustrates that not acute postnatal kidney function decrease but simply low clearance relative to creatinine production is responsible for the initial rise in Scr after birth. Postnatal water volume constriction might be another contributing factor to this transient increase in Scr, although repeated weight measurements in the first days of life in ELBW cases are uncommon in clinical practice.

Interestingly, a high ratio > 1 (indicating higher amount of creatinine produced than eliminated) was observed during the first 3 days despite constant production of creatinine and increase in clearance of creatinine after birth. It may be hypothesized that this observation is linked to reabsorption of creatinine by immature kidney tubules and that both postnatal (time after birth) and prenatal (gestational age) maturation reduce the reabsorption capability of these tubules. Findings (or results) in our study indicate that GA is an additional factor associated with this ratio, meaning that the amount of creatinine that is being reabsorbed differs per GA.

Finally, especially in a population where the blood volume available for sampling is limited and the risk of anemia is high, restrained management should be pursued when collecting blood samples for e.g. Scr determination. The developed reference ranges could help gain insights in normative creatinine profiles in ELBW neonates and assist in the discrimination between whether a change in Scr concentration is either due to developmental physiological aspects (e.g. as a result of GA or increasing PNA) or due to pathophysiological conditions (e.g. asphyxia or sepsis). For instance, an increase in Scr during the first days of life simply reflects that creatinine production is greater than the eliminated amount per day, up to day 2–3 after birth. The decline in Scr afterwards reflects ongoing postnatal maturation. Observing a substantial increase in Scr after birth for an ELBW neonate born at 32 weeks GA could reflect a pathological condition (e.g. acute kidney injury) rather than immaturity of the kidney [30]. Along the same line, the absence of a Scr decrease may reflect impaired kidney function. Although this study is not focusing on acute kidney injury in ELBW neonates, it needs to be acknowledged that physiological events early after birth cause these parameters to change irrespective of acute kidney injury and that neonatal physiology must be considered in a neonatal-specific definition of acute kidney injury [31].

In conclusion, we report the first mathematical model that is able to characterize creatinine dynamics in ELBW neonates during the first 6 weeks of life in a quantitative manner. Three investigated characteristics influence creatinine clearance, namely GA and MOD (influencing maturation of clearance) and treatment with ibuprofen (directly influencing clearance, but not impeding its maturation), while prenatal lung maturation or the need for treatment with inotropic agents were not significant covariates. Gestational age is further the major determinant for the initial creatinine concentration after birth (increasing 6% per week of GA) suggesting gestational-dependent maternal creatinine transfer to the maternal compartment until birth. The model-derived GA-adjusted reference intervals for ELBW neonates provides a rationale for normative Scr concentrations and as such may assist clinicians to further optimize monitoring and treatment decisions in this vulnerable patient population or to assess adverse drug reactions. This study illustrates a physiological approach to establish population-specific reference ranges, as neonatal physiology is distinct from that of older infants, and suggests a methodological approach to attain this.