Risk Prediction Score for Chronic Kidney Disease in Healthy Adults and Adults With Type 2 Diabetes

Systemic Review

Alejandra González-Rocha, MS; Victor A. Colli, MD; Edgar Denova-Gutiérrez, PhD

Prev Chronic Dis. 2023;20(4):E30

Abstract and Introduction

Abstract

Introduction: Chronic kidney disease (CKD) is an important public health problem. In 2017, the global prevalence was estimated at 9.1%. Appropriate tools to predict the risk of developing CKD are necessary to prevent its progression. Type 2 diabetes is a leading cause of CKD; screening the population living with the disease is a cost-effective solution to prevent CKD. The aim of our study was to identify the existing prediction scores and their diagnostic accuracy for detecting CKD in apparently healthy populations and populations with type 2 diabetes.

Methods: We conducted an electronic search in databases, including Medline/PubMed, Embase, Health Evidence, and others. For the inclusion criteria we considered studies with a risk predictive score in healthy populations and populations with type 2 diabetes. We extracted information about the models, variables, and diagnostic accuracy, such as area under the receiver operating characteristic curve (AUC), C statistic, or sensitivity and specificity.

Results: We screened 2,359 records and included 13 studies for healthy population, 7 studies for patients with type 2 diabetes, and 1 for both populations. We identified 12 models for patients with type 2 diabetes; the range of C statistic was from 0.56 to 0.81, and the range of AUC was from 0.71 to 0.83. For healthy populations, we identified 36 models with the range of C statistics from 0.65 to 0.91, and the range of AUC from 0.63 to 0.91.

Conclusion: This review identified models with good discriminatory performance and methodologic quality, but they need more validation in populations other than those studied. This review did not identify risk models with variables comparable between them to enable conducting a meta-analysis.

Introduction

Chronic kidney disease (CKD) has been defined as abnormalities of kidney structure or function present for more than 3 months.^[1] CKD is a public health problem.^[2–4] According to data from the Global Burden of Disease (GBD) study, in 2017^[4] the prevalence of CKD was estimated at 9.1% globally. Of total mortality, 4.6% of deaths were attributable to CKD and cardiovascular disease (CVD), which was attributable to impaired kidney function.

Type 2 diabetes became the second leading cause of CKD and CKD-related deaths in 2019.^[3] Impaired fasting plasma glucose, high blood pressure, high body mass index, a diet high in sodium, and lead were risk factors for CKD quantified in GBD. Approximately 31% of CKD disability-adjusted life years were attributable to diabetes.^[4]

After automatic reporting of the glomerular filtration rate (eGFR) began, referrals to nephrology specialists by primary care services increased. However, the proportion of appropriate referrals did not change, indicating a need to develop appropriate screenings for CKD.^[5] Persons living with hypertension, diabetes, or cardiovascular diseases should be screened for CKD; identifying and treating CKD would reduce the burden of kidney disease.^[6] CKD can be detected early through inexpensive interventions.^[4]

Echouffo-Tcheugui and Kengne presented a systematic review with 30 models predicting the occurrence of CKD and concluded that some models had acceptable discriminatory performance.^[7] CKD screening in groups at high risk is likely to be cost-effective. Predictive models that incorporate clinical information systems would facilitate improved treatment allocations and health care management.^[6,8]

The aim of our study was to identify the existing prediction risk scores and their diagnostic accuracy for detecting CKD in apparently healthy adults and adults living with type 2 diabetes.

1 2 3 4 5

Next Section

Abstract and Introduction
Methods
Results
Discussion
Appendix
References
Sidebar

Table 1. Characteristics of Studies Included in Analysis of Studies of Chronic Kidney Disease (CKD) in Healthy Adults and Adults Living With Type 2 Diabetes
Author and year	Population, total n; outcomes, n (age)	Study design	Type of statistical analysis	Model identification	Variables included	Accuracy predictor
Healthy adults
Al-Shamsi et al (10), 2019	622; 71 (52.4 y)	Retrospective cohort	Fine and gray regression	Full model	Age, sex, diabetes mellitus, hypertension, dyslipidemia, smoking, cardiovascular disease, systolic blood pressure, diastolic blood pressure, total cholesterol, triglycerides, HbA_1c, eGFR	AUC
Al-Shamsi et al (10), 2019	622; 71 (52.4 y)	Retrospective cohort	Fine and gray regression	Stepwise model	eGFR, diabetes, cholesterol, HbA_1c	AUC
Chien et al (11), 2010	5,168; 190 (51.2 y)	Prospective cohort	Cox proportional hazards regression	Clinical model	Age, BMI, diastolic blood pressure (mm Hg), history of type 2 diabetes, history of stroke	C statistic, sensitivity, specificity
Chien et al (11), 2010	5,168; 190 (51.2 y)	Prospective cohort	Cox proportional hazards regression	Biochemical model	Age, diastolic blood pressure, history of stroke, uric acid, postprandial glucose, HbA_1c, urine protein ≥100 mg/dL	C statistic, sensitivity, specificity
Hao et al (12), 2017	1,310,363; 7,448 (NR)	Retrospective cohort	Multivariable logistic regression	Model derivation^a	146 clinical variables including demographics, diagnosis of type 2 diabetes, medications, laboratory test results, and resource utilization	C statistic, sensitivity, specificity
Hao et al (12), 2017	1,430,772; 8,299 (NR)	Retrospective cohort	Multivariable logistic regression	Model validation		C statistic, sensitivity, specificity
Hippisley-Cox and Coupland (13), 2010	1,591,884; 23,786 (35–74 y)	Prospective cohort	Cox proportional hazards regression	Final model	Age, ethnicity, deprivation, smoking, BMI, systolic blood pressure, type 2 diabetes, rheumatoid arthritis, cardiovascular disease, treated hypertension, congestive cardiac failure; peripheral vascular disease, NSAID use, family history of kidney disease, systemic lupus erythematosus (in women), and kidney stones (in women)	AUC
Halbesma et al (14), 2011	6,809; 272 (28–75 y)	Prospective cohort	Multivariable logistic regression	Final model	Age, urinary albumin excretion, systolic blood pressure, C-reactive protein, known hypertension	AUC
Kwon et al (15), 2012	2,921; NR (≥19 y)	Cross-sectional	Multivariable logistic regression	NR	Age, sex, anemia, hypertension, diabetes, cardiovascular disease, and proteinuria	AUC, sensitivity, specificity
Kwon et al (15), 2012	External validation 8,166; NR (≥30 y)	Cross-sectional	Multivariable logistic regression	NR	Age, sex, anemia, hypertension, diabetes, cardiovascular disease, and proteinuria	AUC, sensitivity, specificity
Lee et al (16), 2019	9,080; 734 (51.8 y)	Prospective cohort	Cox proportional hazards regression	Model 1	Sex, BMI, education level, income, fasting plasma glucose, serum albumin	AUC, C statistic
				Model 2	Sex, BMI, education level, income, fasting plasma glucose, serum albumin, Framingham risk score
				Model 3	Sex, BMI, education level, income, fasting plasma glucose, serum albumin, eGFR, proteinuria
				Model 4	Sex, BMI, education level, income, fasting glucose, serum albumin, eGFR, proteinuria, Framingham risk score
Nelson et al (29), 2019	5,222,711; 974,502 (NR)	Cross-sectional	Multivariable logistic regression	Primary model	Age, sex, race, ethnicity, eGFR, history of cardiovascular disease, ever smoker, hypertension, BMI, albuminuria	C statistic
O'Seaghdha et al (17), 2012	2,490; 229 (45–64 y)	Prospective cohort	Multivariable logistic regression	Model 1: clinical model	Age, type 2 diabetes, hypertension	C statistic
				Model 2: clinical model and baseline eGFR	Age, diabetes mellitus, hypertension, baseline eGFR
				Model 3: model 2 plus measure of proteinuria	Age, diabetes mellitus, hypertension, baseline eGFR, quantitative albuminuria (urine ACR or dipstick proteinuria)
Saranburut et al (18), 2017	3,186; 271 (25–54 y)	Prospective cohort	Multivariable logistic regression	Model 1 (clinical)^a	Age, sex, history of diabetes, systolic blood pressure, waist circumference	AUC
				Model 1a^a	Substitution of waist circumference with overweight (BMI ≥25)
				Model 1b^a	Substitution of hypertension for systolic blood pressure
				Model 2 (clinical plus limited laboratory tests)^a	Age, sex, systolic blood pressure, diabetic mellitus, GFR category
				Model 2a^a	Substitution of systolic blood pressure with hypertension
				Model 3 (clinical plus full laboratory tests)^a	Age, sex, systolic blood pressure, diabetic mellitus, GFR, uric acid, hemoglobin
				Model 3a^a	Substitution of hypertension for systolic blood pressure
				Model 1 (clinical)	Age, sex, history of diabetes, systolic blood pressure, waist circumference or BMI
				Model 2 (clinical plus limited laboratory tests)	Age, sex, systolic blood pressure, diabetic mellitus, GFR category
	External validation 1,395 (35–54 y)			Model 1 (clinical)	Age, sex, history of diabetes, systolic blood pressure, waist circumference
	External validation 1,395 (35–54 y)			Model 2 (clinical plus limited laboratory tests)	Age, sex, systolic blood pressure, diabetic mellitus, GFR category
Thakkinstian et al (19), 2011	3,459; 626 (≥18 y)	Cross-sectional	Multivariable logistic regression	Model 1	Age, diabetes, hypertension, history of kidney stones	C statistic
Umesawa et al (20), 2018	58,855; 7,500 (40–74 y)	Prospective cohort	Multivariable logistic regression	Simple risk prediction	Age, eGFR, proteinuria, hematuria	C statistic
	58,855; 7,500 (40–74 y)			Full risk prediction	Age, eGFR, proteinuria, hematuria, BMI, systolic blood pressure, medication for hypertension, glucose tolerance, medication for diabetes mellitus, smoking and alcohol intake
	External validation 76,152; 8,964 (40–74 y)			Simple risk prediction	Age, eGFR, proteinuria, and hematuria
	External validation 76,152; 8,964 (40–74 y)			Full risk prediction	Age, eGFR, proteinuria, hematuria, BMI, systolic blood pressure, medication for hypertension, glucose tolerance, medication for diabetes mellitus, smoking and alcohol intake
Wen et al (21), 2020	3,266; 590 (NR)	Prospective cohort	Multivariable logistic regression	Training: simple clinical model^a	Sex, waist circumference, systolic blood pressure, diabetes mellitus, education	AUC, sensitivity, specificity
				Best fit model^a	Sex, systolic blood pressure, diabetes mellitus, education, triglyceride, urine ACR, C-reactive protein
				Validation: simple clinical model	Sex, waist circumference, systolic blood pressure, diabetes mellitus, education
				Best fit model	Sex, systolic blood pressure, diabetes, education, triglyceride, urine ACR, C-reactive protein
Yu et al (22), 2021	10,049 total; male: 4,117; 157 (NR)	Prospective cohort	Cox proportional hazards regression	Sex-specific CKD male model	eGFR, HbA_1c standard deviation, uric acid, uric acid standard deviation, blood urea nitrogen, albumin, hemoglobin	AUC
Yu et al (22), 2021	10,049 total; female: 5,932; 270 (NR)	Prospective cohort	Cox proportional hazards regression	Sex-specific CKD female model	Age, eGFR, triglycerides, HbA_1c standard deviation, uric acid, uric acid standard deviation, blood urea nitrogen, albumin, hemoglobin, age at menarche (if ≥17 years)	AUC
Adults with type 2 diabetes
Blech et al (23), 2011	1,274; 556 (62.6 y)	Cross-sectional	Multivariable logistic regression	Score 1	Age, duration of diabetes, diabetes type, sex, ethnicity	C statistic, sensitivity, specificity
Blech et al (23), 2011	1,274; 556 (62.6 y)	Cross-sectional	Multivariable logistic regression	Score 2	5 single-nucleotide polymorphisms in 5 genes (HSPG2, NOS3, ADIPOR2, AGER, CCL5), age, duration of diabetes, diabetes type, sex, ethnicity	C statistic, sensitivity, specificity
Dunkler et al (24), 2015	6,766; 1,079 (≥55 y)	Prospective cohort	Multivariable logistic regression	Laboratory model	Urine ACR, eGFR, albuminuria stage (normo- or microalbuminuria), sex, age	C statistic, sensitivity, specificity
Dunkler et al (24), 2015	External validation 8,300 (≥55 y)	Prospective cohort	Multivariable logistic regression	Clinical model	Delta–urine albumin-creatinine ratio to progression, eGFR, albuminuria stage, sex, age, race (White, Asian, other), diabetes duration (years, log transformed), fasting LDL (mg/dL), glucose (mg/dl), waist circumference (cm), comorbidities major atherosclerotic cardiac events (myocardial infarction, stable or unstable angina, coronary artery bypass grafting, or percutaneous interventions, including angioplasty, stenting, atherectomy), laser therapy for diabetic retinopathy, peripheral artery disease (peripheral arterial angioplasty, limb or foot amputation), stroke or transient ischemic attack, number of antihypertensive drugs prescribed	C statistic, sensitivity, specificity
Jardine et al (25), 2012	7,377; 2,715 (NR)	Prospective cohort	Cox proportional hazards regression	eGFR ACR model	eGFR, ACR	C statistic
	7,377; 2,715 (NR)			Final risk prediction model	Ethnicity, eGFR, ACR, systolic blood pressure, hypertension treatment, HbA_1c level, diabetic retinopathy, waist circumference
	External validation 11,140 (NR)			Final risk prediction model	eGFR, urinary ACR, systolic blood pressure, HbA_1c level, diabetic retinopathy, blood pressure–lowering treatment at baseline, Asian ethnicity, waist circumference
Low et al (26), 2017	1,582; 679 (NR)	Prospective cohort	Multivariable logistic regression	Training data set^a	Log urinary ACR (mg/g), systolic blood pressure (per 10 mm Hg), HbA_1c, eGFR, (per 5 ml/min/1.73m²), LDL cholesterol (mmol/L), age (per 10 years increase)	AUC, sensitivity, specificity
Low et al (26), 2017	1,582; 679 (NR)	Prospective cohort	Multivariable logistic regression	Test data set	Log urinary ACR, systolic blood pressure, HbA_1c, eGFR, LDL cholesterol, age (per 10 years increase)	AUC, sensitivity, specificity
Nelson et al (29), 2019	5,222,711; 974,502 (NR)	Cross-sectional	Multivariable logistic regression	Diabetic model	Age, sex, race, ethnicity, eGFR, history of cardiovascular disease, ever smoker, hypertension, BMI, albuminuria, diabetes medications (insulin vs only oral medications vs none), HbA_1c values, and the interaction between diabetes medications and HbA_1c values	C statistic
External validation	2,253,540; 367,159 (NR)			External validation model		C statistic
Raña-Custodio et al (unpublished data)^b	18,148; 1,617 (60.5 y)	Prospective cohort	Multivariable logistic regression	Office equation	Sex, age, BMI, current tobacco smoking and alcohol intake, evolution of diabetes mellitus, current diabetes mellitus treatment scheme (insulin, oral hypoglycemics, or both), prevalent microvascular complication of type 2 diabetes (retinopathy, diabetic foot, neuropathy, or stroke), square of age and female current smoker	C statistic
Raña-Custodio et al (unpublished data)^b	18,148; 1,617 (60.5 y)	Prospective cohort	Multivariable logistic regression	Laboratory risk score	Sex, age, BMI, current tobacco smoking and alcohol intake, family history of CKD (defined as any prevalent history of CKD among any first-degree relative), history of hypertension, evolution of diabetes mellitus, current diabetes mellitus treatment scheme (insulin, oral hypoglycemics, or both), microvascular complication of type 2 diabetes (retinopathy, diabetic foot, neuropathy, or stroke), fasting plasma glucose, HbA_1c, serum creatinine (isotope dilution mass spectrometry), eGFR (Chronic Kidney Disease Epidemiology Collaboration equation), total cholesterol, triglyceride levels	C statistic
Wu et al (27), 2017	4,795; 643 (59.3 y)	Prospective cohort	Multivariable logistic regression	Development model	Sex, BMI, systolic blood pressure, duration of diabetes (years)	AUC
Wysham et al (28), 2020	160,031; 9,973 (NR)	Retrospective cohort	Multivariable logistic regression	DKD	Age, sex, geographic region, insurance type, payer type, adapted diabetes complications severity index, a recorded diagnosis of hypertension, a recorded diagnosis of heart failure, anemia, diabetic nephropathy, CKD stage 1 and 2, time interval with diabetes mellitus	C statistic

Abbreviations: ACR, albumin-to-creatinine ratio; AUC, area under the receiver operating characteristic curve; BMI, body mass index; DKD, diabetic kidney disease; eGFR, (estimated) glomerular filtration rate; HbA_1c, glycated hemoglobin A_1c; LDL, low-density lipoprotein; NR, not reported; NSAID, nonsteroidal anti-inflammatory drug.
^aDevelopment models.
^bA. Raña-Custodio; M. Lajous; E. Denova-Gutiérrez, PhD; M. Chávez-Cárdenas; R. Lopez-Ridaura; G. Danaei, personal communication, 2023.

Table 2. Accuracy Results of Models From Analysis of Studies of Chronic Kidney Disease (CKD) in Healthy Adults and Adults Living With Type 2 Diabetes
Author and year	Model name/stage	AUC (95% CI)	C statistic (95% CI)	Sensitivity	Specificity
Healthy adults
Al-Shamsi et al (10), 2019	Full model	0.90 (0.85–0.95)	NR	NR	NR
Al-Shamsi et al (10), 2019	Stepwise model	0.91 (0.85–0.96)	NR	NR	NR
Chien et al (11), 2010	Clinical model	NR	0.76	0.76	0.66
Chien et al (11), 2010	Biochemical model	NR	0.76	0.88	0.51
Hao et al (12), 2017	Model derivation	NR	0.91	62.61 (95% CI, 61.50−63.71)	97.33 (95% CI, 97.30−97.36)
Hao et al (12), 2017	Model validation	NR	0.87	50.33 (95% CI, 49.25−51.41)	96.60 (95% CI, 96.57−96.63)
Hippisley-Cox and Coupland (13), 2010	Final model (THIN)	Male: 0.87 (0.87–0.88); female: 0.87 (0.87–0.88)	NR	NR	NR
Hippisley-Cox and Coupland (13), 2010	(QResearch)	Male: 0.87 (0.87–0.88); female: 0.87 (0.87–0.88)	NR	NR	NR
Halbesma et al (14), 2011	Final model	0.84 (0.82–0.86)	NR	NR	NR
Kwon et al (15), 2012	NR	0.87 (0.84–0.89)	NR	89.4 (95% CI, 84.4–93.2)	70.6 (95% CI, 68.90–72.30)
Kwon et al (15), 2012	External validation	0.78 (0.76–0.80)	NR	NR	NR
Lee et al (16), 2019	Model 1	0.63 (0.61–0.65)	0.65 (0.63–0.67)	NR	NR
	Model 2	0.69 (0.68–0.72)	0.72 (0.70–0.74)	NR	NR
	Model 3	0.79 (0.78–0.81)	0.81 (0.80–0.83)	NR	NR
	Model 4	0.81 (0.80–0.83)	0.83 (0.82–0.85)	NR	NR
Nelson et al (29), 2019	Primary model	NR	0.87 (0.82–0.90)	NR	NR
Nelson et al (29), 2019	External validation	NR	0.84 (0.83–0.87)	NR	NR
O'Seaghdha et al (17), 2012	Model 1: clinical model	NR	0.79	NR	NR
	Model 2: clinical model and baseline eGFR	NR	0.81	NR	NR
	Model 3: model 2 plus measure of proteinuria	NR	0.81	NR	NR
Saranburut et al (18), 2017	Model 1 (clinical)	0.72 (0.69–0.75)	NR	NR	NR
	Model 1a	0.72 (0.69–0.75)	NR	NR	NR
	Model 1b	0.71 (0.68–0.74)	NR	NR	NR
	Model 2 (clinical and limited laboratory tests)	0.79 (0.76–0.82)	NR	NR	NR
	Model 2a	0.78 (0.76–0.81)	NR	NR	NR
	Model 3 (clinical and full laboratory tests)	0.80 (0.77–0.82)	NR	NR	NR
	Model 3a	0.79 (0.76–0.82)	NR	NR	NR
	Model 1 (clinical)	0.71 (0.68–0.74)	NR	NR	NR
	Model 2 (clinical plus limited laboratory tests)	0.75 (0.72–0.78)	NR	NR	NR
	External validation: model 1 (clinical)	0.66 (0.55–0.78)	NR	NR	NR
	Model 2 (clinical and limited laboratory tests)	0.88 (0.80–0.95)	NR	NR	NR
Thakkinstian et al (19), 2011	Model 1	NR	Derivative 0.77	NR	NR
Thakkinstian et al (19), 2011	Model 2	NR	Validated 0.74	NR	NR
Umesawa et al (20), 2018	Simple risk prediction	NR	Male: 0.82; female: 0.82	NR	NR
	Full risk prediction	NR	Male: 0.82; female: 0.82	NR	NR
	External validation: simple risk prediction	NR	Male: 0.82; female: 0.81	NR	NR
	Full risk prediction	NR	Male: 0.83; female: 0.81	NR	NR
Wen et al (21), 2020	Training: simple clinical model	0.71 (0.68–0.74)	NR	NR	NR
	Training: best-fit model	0.72 (0.69–0.75)	NR	NR	NR
	Validation: simple clinical model	0.71 (0.68–0.74)	NR	70.49 (95% CI, 63.30–77.00)	65.14 (61.90–68.30)
	Validation: best-fit model	0.72 (0.69–0.74)	NR	56.83 (95% CI, 49.30–64.10)	76.61 (73.70–79.40)
Yu et al (22), 2021	Sex-specific CKD male model	0.93 (0.90–0.96)	NR	NR	NR
Yu et al (22), 2021	Sex-specific CKD female model	0.95 (0.93–0.97)	NR	NR	NR
Adults living with type 2 diabetes
Blech et al (23), 2011	Score 1	NR	0.56	64.55	46.54
Dunkler et al (24), 2015	Laboratory model	NR	0.67	NR	NR
	Clinical model	NR	0.69	NR	NR
	External validation	NR	0.69	NR	NR
Jardine et al (25), 2012	eGFR plus albumin-to-creatinine ratio model	NR	0.62 (0.61–0.64)	NR	NR
	Final risk prediction model	NR	0.64 (0.63–0.65)	NR	NR
	External validation	NR	0.64	NR	NR
	External validation	NR	0.62	NR	NR
Low et al (26), 2017	Training data set	0.80 (0.77–0.83)	NR	71.4	72.2
Low et al (26), 2017	Test data set	0.83 (0.79–0.87)	NR	75.6	72.3
Nelson et al (29), 2019	Diabetic model	NR	0.80 (0.79–0.83)	NR	NR
Nelson et al (29), 2019	External validation	NR	0.81 (0.80–0.82)	NR	NR
Raña-Custodio et al (unpublished data)^a	Office equation	NR	0.67	NR	NR
Raña-Custodio et al (unpublished data)^a	Laboratory risk score	NR	0.71	NR	NR
Wu et al (27), 2017	Development model	0.71 (0.69–0.73)	NR	NR	NR
Wysham et al (28), 2020	DKD model	NR	0.70	NR	NR

Abbreviations: AUC, area under the receiver operating characteristic curve; CKD, chronic kidney disease; DKD, diabetic kidney disease; NR, not reported.
^aA. Raña-Custodio, M. Lajous, E. Denova-Gutiérrez, PhD, M. Chávez-Cárdenas, R. Lopez-Ridaura, G. Danaei, personal communication, 2023.