In sickle cell disease (SCD), reduced estimated glomerular filtration rate (eGFR) and albuminuria are frequently seen and associated with increased mortality in SCD.¹ In non-SCD patients, a decline of renal function over time is a risk factor for cardiovascular morbidity and mortality.² This correlation between renal function decline and cardiovascular outcomes might be explained by the fact that changes in renal function over time are a reflection of cumulative microvascular damage. The sequence in which nephropathy occurs in SCD (hyperfiltration followed by proteinuria and renal failure) is similar to that in patients with cardiovascular conditions such as hypertension. We hypothesized that the slope of renal function decline may be a valuable representation of vasculopathy in SCD. As this slope is supposedly constant throughout life, it might be a valuable biomarker of disease severity even before the onset of overt organ damage, which could help guide treatment decisions (eg, selecting the highest-risk patients for curative treatment options such as hematopoietic stem cell transplantation).

For this longitudinal cohort study, electronic records from consecutive adult SCD patients (HbSS, HbSβ-thalassemia, HbSC) visiting the outpatient clinic of the Amsterdam University Medical Centers between 2006–2019, were reviewed. The use of ethnicity in eGFR equations is currently under debate.³ Additionally, aging, representing an important variable of eGFR equations (just as ethnicity and sex), will by definition lead to renal function decline in follow-up studies. Therefore steady-state plasma creatinine values were used to determine the gradual change in renal function over time. For every plasma creatinine level, a reciprocal creatinine level was calculated (1/plasma creatinine in mg/dL) as this displays a linear relationship over time that is not influenced by ethnicity, sex, or age.⁴ All reciprocal creatinine values for each patient were subjected to a mixed model regression analysis for repeated measures, to calculate individual directional coefficients resembling the slope of renal function. Patients with <3 creatinine values were excluded. Patients were considered lost to follow-up if they had not visited the hospital for >5 years at the end of follow-up.

To investigate the association between the slope of renal function decline and organ damage development, a mixed model regression analysis with a random slope and random intercept was performed by including an interaction term between the slope and the variable of interest. Associations between the slope and laboratory values that were previously linked to mortality (ie, hemolysis and N-terminal-pro-brain natriuretic peptide levels (NT-proBNP)) were investigated, again by using mixed model regression analysis.⁵

Of 203 eligible patients, 169 patients with 2147 creatinine values were included (Figure S1). The median age at the end of follow-up was 34 years [IQR 27–44], 52% were female at birth and 63% had a severe genotype (HbSS/HbSβ⁰) (Table S1). Fifty-six (33%) patients had chronic kidney disease, while hypertension and diabetes were observed in 18 (11%) and 3 (2%) patients, respectively. Hydroxyurea was prescribed in 53 (29%) patients, and 24 (9%) patients received periodic exchange transfusions. Sixty-six (39%) patients experienced frequent vaso-occlusive crisis (VOCs). Median creatinine values were 0.69 mg/dL [0.54–0.8] at the beginning of the study and 0.79 mg/dL [0.62–0.96] at the end of follow-up.

Patients were divided into tertiles based on their slope of renal function decline (Table S2 and Figure S2). Tertile 1 represents patients with the steepest decline in renal function and tertile 3 patients with the lowest decline. Age, sex at birth, and follow-up time were equally distributed between tertiles. Tertile 1 contained a larger proportion of patients with a severe genotype (HbSS/HbSβ⁰) compared to tertiles 2 and 3 (84%, 60%, and 54%, respectively, p < .001). Tertile 1 also consisted of the largest proportion of patients with frequent VOCs (57%, 30%, and 30%, respectively, p = .007), hydroxyurea treatment (45%, 26%, and 23%, respectively, p = .03) and periodic exchange transfusions (25%, 9%, 9%, respectively, p = .03). The hemoglobin concentration was lowest in tertile 1 (9.0 g/dL [7.4–10.3] versus 9.9 g/dL [8.5–11.0] in tertile 2 and 10.2 g/dL [9.0–11.4] in tertile 3, p = .003), whereas plasma lactate dehydrogenase (LDH) concentration was highest in tertile 1 (410 U/L [312–524] versus 321 U/L [242–424] and 323 U/L [234–410] in tertiles 2 and 3, respectively, p = <.001). Additionally, plasma NT-proBNP levels were significantly higher in tertile 1 (92 ng/L [50–160] versus 56 ng/L [50–148] and 50 ng/L [49–73] in tertiles 2 and 3 respectively, p = .01).

The occurrence of organ damage and mortality was presented as prevalence and incidence per 1000 person-years (Tables S3 and S4). Twelve patients died during follow-up (mortality rate: 7%), of which five were in tertile 1, four in tertile 2, and three in tertile 3. Albuminuria prevalence was significantly different between the tertiles (48%, 23%, and 29%, respectively, p = .001). History of acute chest syndrome (ACS) was more frequent in tertile 1 compared to tertiles 2 and 3 (45%, 26%, and 32%, respectively, p = .006).

In mixed model regression analyses adjusted for genotype and age, there was a significant association between the slope of renal function decline and albuminuria and chronic kidney disease (CKD) (adjusted-OR 1.10, p = .005) and history of ACS (adjusted-OR 1.11 p = .001; Table 1). Although patients in tertile 1 were more likely to have frequent VOCs, frequent VOCs were not significantly associated with renal function decline in mixed model regression analyses corrected for genotype (data not shown).

There was a significant association between the slope of renal function decline and cerebrovascular accident (CVA) as a composite endpoint with mortality (OR 1.10, p = .045) and a trend was found for AVN as a composite endpoint with mortality (OR 1.07, p = .06 (Table S5)). Regarding laboratory parameters, there were significant associations between the slope of renal function decline and hemoglobin levels (adjusted-OR 0.97, p = .02) and plasma LDH levels (adjusted-OR 1.0002, p = .02). Also, adjusted for genotype, hemoglobin, and age, there was a significant association between the renal slope and plasma NT-proBNP level (adjusted-OR 1.0004, p = .03), Table 1. A sensitivity analysis with only the severe genotypes showed comparable results (data not shown).

Although outcomes of SCD patients have greatly improved over the past decades and curation is possible, curative treatments come with risks of serious adverse events. Selecting the highest-risk patients who could benefit most from such intensive treatment would be valuable, but due to the remarkable clinical variability within SCD patients, disease severity can differ significantly. The slope of renal function decline of an individual is easily accessible and interpretable, is stable over time, and we confirmed its linearity in SCD patients. Therefore, this marker could be useful in early risk assessment. As we observed a significant interaction of the renal slope with parameters of severe disease, such as a history of ACS, hemoglobin levels, parameters of hemolysis (e.g., LDH levels), and NT-proBNP plasma levels, this might indicate an association between disease severity and the slope of renal function decline. Even after adjusting for genotype, these associations remained significant, indicating that the association with disease severity is not solely driven by more severe genotypes (HbSS/HbSβ⁰) in the first tertile. Furthermore, we observed a trend towards a higher mortality rate in the first tertile, but there was no significant interaction between the slope of renal function decline and mortality in mixed model regression analysis. This could be due to the relatively low number of deaths in our study since previous studies have repeatedly reported correlations between a decrease in eGFR and mortality in SCD.¹ Nevertheless, plasma NT-proBNP, a known risk factor for mortality in SCD, was significantly associated with the slope of renal function decline, suggesting that with longer follow-up, or in cohorts with more heavily affected patients, differences in mortality rate may appear.

The strengths of this study are the large patient cohort and long follow-up of 13.5 years. In addition, we used reciprocal plasma creatinine values to evaluate renal function decline over time instead of eGFR formulas that correct for ethnicity (in line with recommendations of renal societies³), age, and sex. The observations described in this manuscript should be considered in the context of the limitations associated with the retrospective single-center design of this study. Due to patients who were lost to follow-up and missed appointments, there were missing data throughout the observation period. This was partially addressed by excluding patients with <3 steady-state plasma creatinine measurements and those who were lost to follow-up for >5 years. Although kidney function has consistently been shown to change linearly over time, also in SCD patients,⁶ future studies are warranted to determine the minimum amount of creatinine values necessary to calculate a reliable individual linear line. Additionally, the effect of hyperfiltration on the linearity of the slope is unknown and this may impede the use of this slope in pediatric patients.

In conclusion, this study demonstrates that the slope of renal function decline is associated with multiple indicators of disease severity in SCD. Given the linear decline over time, this slope might be a valuable biomarker that may contribute to the early identification of patients at the highest risk and might contribute to personalized treatment decisions. Larger studies are needed to confirm and validate these findings and assess whether the slope of renal function decline predicts disease severity early in life.

FUNDING INFORMATION

The authors received no specific funding for this work.

CONFLICT OF INTEREST STATEMENT

The authors declare that there is no conflict of interest.