Acute kidney injury in hospitalized children with sickle cell anemia

Between January and August in 2019, children admitted to Mulago National Referral Hospital with SCA and VOC were screened for eligibility. Eligible children were prospectively recruited in a hospital-based study to define AKI prevalence, risk factors, and outcomes. Inclusion criteria were hospital admission, documented SCA by hemoglobin electrophoresis, age 2 to 18 years, a pain score ≥ 2 using an age-specific pain scale, and a willingness to complete study procedures. Pain in children aged 2–3 years was assessed using the face, legs, activity, cry, and consolability (FLACC) scale [12], the Wong-Baker Faces pain scale in children 3–7 years of age, and the numeric pain scale in children ≥ 8 years [13].

Children were recruited from the paediatric emergency unit that admits an average of 3 children daily with SCA and VOC. Most of the children admitted with VOC also attend the hospital's dedicated sickle cell clinic that provides medical care to approximately 1400 children per year. Routine care includes malaria prophylaxis using sulfadoxine-pyrimethamine, daily folic acid, and penicillin V for children < 5 years of age. Only a small proportion of children receive hydroxyurea.

The sample size was calculated using the formula \(n=\left({z}^{2}p\left(1-p\right)/{d}^{2}\right)\), where; z = standard normal variate corresponding to the 95% confidence interval and is 1.96, d = the required precision of the estimate (0.05), p = prevalence rate [14]. The sample size was calculated based on previous studies indicating AKI prevalence of 2.3%, 6.9%, and 13% among children with SCA and VOC [6], generating a sample size of 173.

After informed consent and enrolment, all children had a complete history and physical examination conducted by a study doctor to assess medication use, signs of infection, the site and severity of pain. Blood pressure was calculated as the mean of three independent measurements and hypertension was defined as a systolic blood pressure > 95% percentile and/or a diastolic blood pressure > 95% for children < 13 years of age or a systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 80 mmHg for children 13 years of age or older [15]. Liver size was evaluated by abdominal palpation below the subcostal margin and defined as: i) no hepatomegaly, liver not palpable; ii) hepatomegaly, liver palpable below the costal margin; iii) tender hepatomegaly, tenderness on palpation of the liver. A spot urine sample was collected from each recruited child for dipstick urinalysis, urine microscopy, assessment of urine albumin and urine creatinine. Microalbuminuria was defined as urine albumin-to-creatinine ratio of 3-30 mg/mmol and macroalbuminuria as urine albumin-to-creatinine ratio, > 30 mg/mmol.

Children were assessed for symptoms, signs of infection and had a complete blood count and malaria evaluation by Giemsa microscopy. Sepsis was defined using International paediatric sepsis consensus guidelines [16] based on two or more of the following criteria, one of which must be abnormal temperature > 38.5 °C or < 36 °C: age-specific tachycardia, age-specific tachypnea, or leukocytosis. We used age-specific thresholds for leukocytosis in children with SCA [17]. We categorised children as having hypovolemia if they were unable to drink or breastfeed, had diarrhea or vomiting, were hypotensive, or capillary refill time > 3 s (features of dehydration). Children were diagnosed with a urinary tract infection based on a positive nitrite or leukocyte test by urinalysis in a child with a history of fever. An acute infection was defined by the presence of sepsis, malaria or urinary tract infection. Microbiology to culture blood or urine samples was not feasible.

Cystatin C— a biomarker of glomerular filtration [6] — was measured on cryopreserved serum using an ELISA correlated to the reference standard (Catalog # ERM-DA471/IFCC) with a slope of 1.07 and R² value of 0.998 (Quantikine® immunoassay, R&D Systems, Minneapolis, MN) [18]. The assay has been evaluated for matrix effects using samples from apparently healthy donors with a mean (SD) of 0.79 mg/L (1.6) for serum and 0.77 mg/L (1.6) for EDTA plasma [18]. On each plate pooled serum samples were tested in duplicate alongside a commercial control (QC23: Quantikine Immunoassay Control Group 8, R&D Systems, Minneapolis, MN). Laboratory technicians were blinded to participant details.

AKI was defined based on the Kidney Disease: Improving Global Outcomes (KDIGO) guidelines as an increase in SCr of ≥ 0.3 mg/dl within 48 h or a 50% increase in baseline creatinine within 7 days [10]. Daily 24 h urine output was not quantified over hospitalization. SCr was assessed in whole blood on enrolment, at 24–48 h, and day 7 or discharge (whichever happened earlier) using iSTAT CHEM8 + cartridges (Abbott Point of Care Inc., Princeton, NJ). This approach of SCr measurement uses an enzymatic assay with values traceable to the U.S. National Institute of Standards and Technology (NIST) standard reference material SRM909, with a reportable range of 0.20–20.0 mg/dL. To define AKI, SCr values below the reportable range were assigned a value of 0.19 mg/dL. Each participant’s lowest measured creatinine, or the nadir, was taken as the baseline SCr for AKI determination. None of the patients had a known baseline creatinine prior to hospitalization. As SCA is a risk factor for kidney disease, population-based estimates of baseline serum creatinine would be inaccurate in this patient population. In instances where a child had only a single creatinine measure (n = 7), the Pottel height-independent GFR estimating equation was used to back-calculate baseline creatinine, assuming a normal GFR of 120 mL/min per 1.73m² [19, 20].

Using the KDIGO guidelines, AKI was staged based on creatinine fold change from the nadir to the maximum, where: stage 1 included a 1.5- <2.0-fold change in creatinine from baseline or a ≥ 0.3 mg/dl increase in creatinine within 48 h; stage 2, 2– <3.0-fold change in serum creatinine from baseline; stage 3, ≥3.0-fold change in creatinine from baseline, or an increase in creatinine to ≥ 4.0 mg/dl, or an eGFR ≤ 35 ml/min per 1.73 m². AKI was assessed in two ways: using the original KDIGO definition and using a sickle cell anemia modified KDIGO definition (sKDIGO). The point-of-care i-STAT system reports creatinine in 0.1 mg/dL increments so a one-unit change in creatinine from 0.2 to 0.3 mg/L constituted AKI using the KDIGO definition. In the sKDIGO definition, children with a nadir creatinine of 0.2 mg/dL with 1.5-fold increase to 0.3 mg/dL were excluded from the AKI definition. The eGFR was calculated using the CKiD creatinine and Cystatin C based formulae [21]: creatinine, eGFR = (0.413*height)/SCr; Cystatin C, eGFR = 70.69 *(Serum Cystatin C)^−0.931; SCr & Cystatin C, eGFR = 39.8*((height/SCr)^0.456)*((1.8/Cystatin C)^0.418)* ((30/bun) ^0.079)*(1.076^male) *((height/1.4)^0.179).

A total of 185 children with SCA and VOC were enrolled into the study (Fig. 1). The median age was 8.9 years (IQR, 5.9—11.8) and 41.6% of participants were female (Table 1). The prevalence of undernutrition was high with 30.8% of children stunted (height-for-age z score < –2), 25.9% underweight (weight-for-age z score < -2), and 29.7% wasted (weight-for-height z score or BMI-for-age z score < -2). Of the 185 children enrolled, 24 (13.0%) children had a history of stroke, and 78 (42.2%) had been hospitalized in the previous six months (Table 1). The majority (70.3%) of children had severe anemia with a hemoglobin < 8 g/dL and 38 (20.5%) children had evidence of an acute infection defined as sepsis (n = 19, 10.3%), malaria (n = 17, 9.6%), or urinary tract infection (n = 4, 2.2%). Six (3.2%) of the participants died during hospitalization and all deaths occurred in female participants (mortality, 7.8%).

The median duration of pain prior to admission was 3 days. The reported locations of pain included the lower limbs (64.9%), abdomen (41.6%), upper limbs (30.3%), back (29.7%), chest (34.6%) and others (3.8%). Analgesia use for children included paracetamol (93.0%), morphine (86.0%), ibuprofen (83.8%), diclophenac (10.3%), codeine (2.2%) and tramadol (1.1%).

Overall, 90/185 (48.7%) children had AKI using the KDIGO definition with the majority of AKI cases occurring on admission 61/90 (67.8%) while 34/90 (37.8%) children had incident or worsening AKI over hospitalization (Fig. 2). In children with AKI, 23/90 (25.6%) of cases occurred in children with a baseline creatinine of 0.2 mg/dL that increased to 0.3 mg/dL over hospitalization. Using the sKDIGO definition, 67/185 (36.2%) children had AKI with 26 cases diagnosed based on a 0.3 mg/dL increase in creatinine within 48 h and 41 cases based on a fold increase in creatinine over 7 days; 43/67(64.2%) had AKI on admission and 28/67 (41.8%) developed incident or worsening AKI during hospitalization (Fig. 2). There were no relationships between medication use and AKI (p > 0.05 for all).

We evaluated the relationship between both AKI definitions and clinical chemistries and urinalysis results (Table 2). Overall, differences in laboratory parameters were larger in children who had AKI defined using sKDIGO definition compared to KDIGO (Table 2). We further evaluated the KDIGO and sKDIGO definition using Cystatin C, an alternative marker of GFR (Fig. 3). Cystatin C was only elevated in children meeting the sKDIGO definition of AKI (p < 0.001).

Finally, we compared outcomes in children based on AKI status. AKI was associated with increased mortality using sKDIGO (mortality, 0.9% no AKI vs. 7.5% AKI, p = 0.024) but not the original-KDIGO definition (mortality, no AKI, 1.1% vs. 5.6% AKI, p = 0.110). AKI on admission was not associated with increased mortality. Rather, children who developed incident AKI in-hospital or those with worsening AKI had increased mortality (p < 0.05 for both). As the sKDIGO definitions was a better predictor of mortality, subsequent analyses evaluating risk factors for AKI and AKI progression are conducted using sKDIGO.

In order to improve recognition of AKI in children with SCA who require hospitalization, we evaluated differences in the demographic characteristics, signs and symptoms on admission and laboratory findings in children with AKI (Table 3). As 43/67 (64.2%) children had AKI on admission to hospital, we are unable to attribute causality to these risk factors. However, understanding risk factors in this high-risk population may enable clinicians to identify features that raise their index of suspicion minimizing exposure to potential nephrotoxic agents and prioritizing monitoring in children at highest risk of AKI. Using logistic regression, AKI risk factors included age (aOR, 1.10, 95% CI 1.01, 1.20), hypovolemia (aOR, 2.98, 95% CI 1.08, 8.24), tender hepatomegaly (aOR, 2.46, 95% CI 1.05, 5.77), and an acute infection (aOR, 2.63, 95% CI 1.19, 5.81) (p < 0.05).

To determine risk factors in children with community-acquired AKI versus hospital-acquired or worsening AKI, we stratified analyses based on the timing of diagnosis (Table 4). Children with community-acquired AKI were older, were more likely to have tea coloured urine, and hepatomegaly or tender hepatomegaly. In addition, AKI on admission was associated with hypovolemia and an acute infection and laboratory findings consistent with an acute inflammatory event. In contrast, participants with incident or worsening AKI over hospitalization were similar in age, with comparable frequencies of infection and similar laboratory findings. There were no differences in nephrotoxic medications use in children with incident or worsening AKI (Table 4). Participants who developed incident or worsening AKI were more likely to be female.