Introduction

The Sickle Cell Outcomes Grading System (SCOGS) is a disease severity grading system developed to standardize disease complications, inform endpoints for clinical research and create definitions for disease manifestations. SCOGS was developed by a panel of adult and pediatric providers using a 3-round Delphi methodology and includes definitions, diagnostic criteria, severity gradings (1-5), and frequency classifications (acute, chronic, or chronic with exacerbation) for 53 SCD related health outcomes. For Chronic kidney disease (CKD), a common complication in SCD, SCOGS adapted the 2024 Kidney Disease: Improving Global Outcomes (KDIGO) definition for CKD and classified severity scoring into to a 5-point severity scale.

We sought to investigate the feasibility of applying SCOGS severity scoring of CKD in two independent datasets.

Methods

We extracted variables from two existing registries and tested the feasibility of applying SCOGS CKD classification, the Globin Regional Network Database and Development (GRNDaD) and the Sickle Cell Clinical Research and Intervention Program (SCCRIP). Both registries are multi-center, large, longitudinal SCD cohorts (>1500 participants each), including pediatric and adult patients with SCD of all genotypes (Lanzkron et al., 2022; Hankins et al., 2018). Each data set was evaluated using the SOCGS CKD classification, assessing availability of variables necessary for scoring, percentages of cohort able to be scored (patient and event level), demographics, ease of data abstraction, and ease of scoring.

SCOGS defines CKD as abnormalities of kidney structure or function, present for >3 months, either by markers of kidney damage such as albuminuria or by decreased Glomerular Filtration Rate (GFR). CKD grades are determined by combining eGFR and albuminuria levels, with grade 1 representing mild kidney impairment and grade 5 indicating end-stage kidney disease or death caused by CKD. Inclusion criteria required at least two measurements of estimated eGFR and/or microalbumin, taken at least three months apart, and the maximum severity score observed was used for classification. All analyses were performed using R, with version 4.5.1 used for SCCRIP and version 4.4.1 used for GRNDaD.

Results

Of 1,382 eligible patients in SCCRIP, 853 (61.7%) met inclusion criteria after data cleaning, ensuring measurements of height and microalbumin were within 365 days of serum creatinine levels. Of the 853 patients, 182 (21.3%) were able to be scored. The median age of scored patients with CKD in the SCCRIP cohort was 13.2 years, with 72.9% aged <18 years. The cohort comprised 60.7% patients with sickle cell anemia genotypes (Hgb SS/Sβ0thalassemia) and 62.4% male patients. The distribution of CKD grades was as follows: 89 (49.2%) patients with Grade 1, 65 (36.0%) with Grade 2, 9 (5.0%) with Grade 3, 11 (6.1%) with Grade 4, and 7 (3.9%) with Grade 5. Of the 2872 patients with complete data for required variables in the GRNDaD cohort, 1339 (46.6%) met inclusion criteria. Among these 1339 patients, 214 (16%) were scored, representing 7.4% of the entire GRNDaD cohort (2,874 patients). Their median age of was 45 years, with only 12 patients (5.6%) being <18 years of age. The cohort comprised 62.1% Hgb SS/Sβ0thalassemia and 37.4% were male patients. The distribution of CKD grades was as follows: 58 (27.1%) patients with grade 1, 90 (42.1%) with grade 2, 16 (7.5%) with grade 3, 17 (7.9%) with grade 4, and 6 (2.8%) with grade 5.

In both cohorts, the primary reason patients with CKD could not be scored was the absence of both microalbumin and eGFR measurements within 365 days of the assessment, with only one of these values available. Aside from missing variables, SCOGS was simple and intuitive to use without and difficulties.

Conclusions

Utilization of SCOGS for CKD was feasible to use in two large, independent cohorts of SCD patients, but with a low proportion of scorable patients. The distribution of CKD severity criteria was similar across both cohorts. The greatest barriers to its use were the unavailability of some variables, particularly microalbumin. Machine learning algorithms could potentially be employed to grade more encounters with missing data. Both cohorts required manual abstraction efforts in addition to automated data pulls. While utilization of SCOGS is feasible in existing datasets, results must be interpreted with caution due to the small number of scored patients.

This content is only available as a PDF.
2025
Sign in via your Institution