The risk evaluation of chronic kidney disease (CKD) patients is crucial for appropriate treatment. The association rule, which is a transparent machine learning approach, aims to share information about CKD risks among diabetic patients; however, its findings are limited to clinical data.
In a recent study published in Scientific Reports, researchers use the association rule to assess CKD risk in the General and Worker diabetes populations.
Study: A comprehensive risk factor analysis using association rules in people with diabetic kidney disease. Image Credit: crystal light / Shutterstock.com
About the study
The lack of risk variables was investigated for its relationship with steady and deteriorating renal function. As an indication of association, a confidence value of more than one was regarded as significant. Individuals were analyzed based on the KidneyDisease: Improving Global Outcomes (KDIGO) CKD risk categories, with albuminuria substituted with dipstick proteinuria.
The study included two cohorts including General and Worker groups. The General cohort was a general diabetic population of 4,935 adults with a mean age of 67 who underwent yearly health assessments in the Kanazawa city and Ishikawa Prefecture of Japan from 1999 to 2018. Adults over the age of 40 without insurance coverage were included in the General cohort.
The Worker cohort included 2,153 insured diabetic workers with a mean age of 48 years who worked for a Japanese firm. Data on these works were obtained from yearly health tests conducted between 2010 and 2016.
Individuals were included in the analysis if their serological creatinine levels were measured and followed up with one or times. Those who were unwilling to participate, lacking risk factors and estimated glomerular filtration rate (eGFR) information at baseline, and could not be followed for five years were removed from the study. The baseline was defined as the initial year in which eGFR was measured.
Diabetes was diagnosed based on fasting blood glucose exceeding seven mmol/L, glycated hemoglobin (HbA1c) of 6.50% or more, or diabetes treatment. The assessed risk factors included urinary protein, eGFR, hemoglobin, HbA1c, alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), triglyceride, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, total cholesterol, diastolic blood pressure, systolic blood pressure, diabetic retinopathy, body mass index (BMI), and a two-year reduction in eGFR from the baseline.
As a proxy measure for end-stage renal disease, a 30% decline in eGFR during the follow-up was considered the study endpoint of decreasing kidney function. The Worker cohort definition included end-stage renal disease diagnosis, kidney transplantation, and commencement of maintenance hemodialysis.
A thorough investigation was performed to investigate the risk factors’ absence and preserved renal function, as well as the risk factors’ presence and decreasing renal function. Furthermore, a stratified analysis was conducted, with the background used for stratification being CKD risk groups, eGFR reduction rate over two years, and diabetic retinopathy.
Good glycemic management was significantly associated with preserved renal function in the General cohort's low-risk categories and Worker cohort's very-high-risk categories with confidence intervals of 0.8 for both groups. Poor glycemic management and decreasing renal function yielded similar findings, with HbA1c confidence levels of 0.4 and 0.3, respectively.
Similarly, in the low-risk General and very-high-risk Worker groups, anemia, obesity, and hypertension all exhibited significant associations. The mean eGFR in the General cohort was 75 mL/min/1.7 m2, 12% of whom had proteinuria. The mean eGFR in the Worker cohort was 80 mL/min/1.7 m2, with several people exhibiting retained renal function.
Over five years, 237 and 110 people in the General and Worker cohorts, respectively, experienced decreasing renal function. To confirm the robustness of the study findings, the researchers assigned risk factor value thresholds to the lowermost or uppermost 10% of the sampled population based on sex, as they had done for the 20% threshold and subsequently categorized the absence or presence of risk factors based on the thresholds. Similar results were obtained for the primary analysis.
Stable kidney function was related to good hyperglycemia and blood pressure management, which is consistent with prior cohort studies and treatment guidelines. The conclusion that LDL cholesterol was not linked with prognosis is consistent with previous observational studies in non-dialysis individuals.
Triglycerides, which were associated to renal events, have also been linked to CKD progression and cardiovascular events. The results of the General cohort revealed that liver injury was associated with decreasing kidney function, thus indicating that liver injury may increase the risk of advanced diabetic renal disease.
The study findings highlight the connections between risk variables and renal events in two diabetes cohorts using association rules. Many of these findings were consistent with prior publications, with the stratified analysis allowing for judgments based on risk categories and retinopathy.
Neverthless, further research in the form of validation studies are needed in cohorts with similar backgrounds.
- Toyama, T., Shimizu, M., Yamaguchi, T. et al. (2023). A comprehensive risk factor analysis using association rules in people with diabetic kidney disease. Scientific Reports 13. doi:10.1038/s41598-023-38811-5