Developing screening tools to estimate the risk of diabetic kidney disease in patients with type 2 diabetes mellitus

Abstract

BACKGROUND:

Diabetic kidney disease (DKD) is an important microvascular complication of diabetes mellitus (DM).

OBJECTIVE:

This study aimed to develop predictive nomograms to estimate the risk of DKD in patients with type 2 diabetes mellitus (T2DM).

METHODS:

The medical records of patients with T2DM in our hospital from March 2022 to March 2023 were retrospectively reviewed. The enrolled patients were randomly selected for training and validation sets in a 7:3 ratio. The models for predicting risk of DKD were virtualized by the nomograms using logistic regression analysis.

RESULTS:

Among the enrolled 597 patients, 418 were assigned to the training set, while 179 were assigned to the validation set. Using the predictors included glycated hemoglobin A1c (HbA1c), high density lipoprotein cholesterol (HDL-C), presence of diabetic retinopathy (DR) and duration of diabetes (DD), we constructed a full model (model 1) for predicting DKD. And using the laboratory indexes of HbA1c, HDL-C, and cystatin C (Cys-C), we developed a laboratory-based model (model 2). The C-indexes were 0.897 for model 1 and 0.867 for model 2, respectively. The calibration curves demonstrated a good agreement between prediction and observation in the two models. The decision curve analysis (DCA) curves showed that the two models achieved a net benefit across all threshold probabilities.

CONCLUSION:

We successfully constructed two prediction models to evaluate the risk of DKD in patients with T2DM. The two models exhibited good predictive performance and could be recommended for DKD screening and early detection.

Keywords

Diabetic kidney disease type 2 diabetes mellitus nomogram risk factor glycated hemoglobin A1c

1. Introduction

Diabetic kidney disease (DKD), which often occurs in patients with diabetes mellitus (DM) without long-term adequate glycemic control, is an important microvascular complication of DM [1]. It has been reported that DKD occurs in approximately 40% of people with type 2 diabetes mellitus (T2DM) and 30% of those with type 1 diabetes mellitus (T1DM) [2]. And the number of DKD patients is expected to increase by nearly 50% over the next 24 years [2]. The most common outcome for DKD patients is renal failure, but the cardiovascular disease and premature mortality also occur frequently [3]. DKD is closely associated with the prognosis of DM patients. Early diagnosis of DKD and appropriate intervention is important to delay the progression of kidney function decline and prevent end-stage kidney disease (ESKD) [3, 4].

For DKD diagnosis, kidney biopsy is the gold-standard approach. However, it is an invasive procedure and cannot be feasible for all patients [5]. It has been reported that several clinical characteristics were correlated with DKD in T2DM patients [6]. There are some predictive models for prediction of incident DKD and meta-analyses had been performed to review the performances of these models [7, 8]. Different prediction models have different concerns, and it is still necessary to develop more prediction models for DKD events. Nomograms are statistical predictive models that incorporate independent factors to estimate prognosis and risk of chronic diseases [9]. In the present study, we aimed to develop predictive nomograms to estimate the risk of DKD in patients with T2DM in the hope that the models would be useful for DKD screening.

2. Methods

2.1 Patients

This retrospective study was approved by the Ethics Committee of our hospital in accordance with the Declaration of Helsinki. Due to the nature of retrospective study and anonymized patient’s information, informed consent was waived.

All patients were inpatients in the endocrine department from March 2022 to March 2023. Patient information was queried and collected by the HIS system from April to May 2023. The inclusion criteria were: (1) patients with T2DM; (2) aged 18–80 years. The main exclusion criteria included: (1) patients with severe systemic disease such as severe infectious diseases (Pestis and cholera), autoimmune diseases, and cancer; (2) patients with severe psychiatric disturbance; (3) pregnant women; (4) patients without complete data.

Generally speaking, the sample size of the study is 10–20 times the total number of factors to be analyzed. Based on this calculation, the sample size for this study was 170–340. In order to obtain a more reliable outcome, we included as many cases as possible.

A training set and a validation set were randomly generated from the study population in a 7:3 ratio. The training set was used to train the prediction models. The validation data were used to validate the models.

2.2 Data collection

DKD was diagnosed as eGFR $<$ 60 mL/min/1.73 m² and/or urine albumin-to-creatinine ratio (UACR) $>$ 30 mg/g for $\geqslant$ 3 months caused by DM [10]. The medical records of enrolled participants were retrospectively reviewed. Factors enrolled for analysis included gender, age, height, weight, body mass index (BMI), fasting plasma glucose (FPG), HbA1c levels, serum creatinine, triglyceride (TG), total cholesterol (TC), high density lipoprotein cholesterol (HDL-C), low density lipoprotein cholesterol (LDL-C), blood urea nitrogen (BUN), cystatin C (Cys-C), presence of hypertension, presence of DR, presence of diabetic foot, presence of diabetic peripheral neuropathy (DPN) and DD. Serum creatinine was detected using isotope dilution liquid chromatography tandem mass spectrometry and can be traced back to IDMS. The T2DM patients with DKD were assigned to the DKD group and the remaining patients were assigned to the non-DKD group.

2.3 Predicted outcomes

The included outcomes for external validation were microalbuminuria, macroalbuminuria, diabetic kidney disease, and chronic kidney disease. The outcomes defined was as follows: DKD was diagnosed as eGFR $<$ 60 mL/min/1.73 m² and/or urine albumin-to-creatinine ratio (UACR) $>$ 30 mg/g for $\geqslant$ 3 months caused by DM Cases with missing data were excluded (83 cases).

2.4 Statistical methods

The data analysis was performed using R4.0.3 software. Whether the continuous data conformed to a normal distribution was tested by the Kolmogorov-Smirnov test. Continuous data were expressed as mean $\pm$ standard deviation (SD) for normal distribution or median (interquartile range) for non-normal distribution. The qualitative data were expressed as absolute values and percentages. Baseline characteristics were compared by Student’s $t$ -test for normally distributed continuous data, Mann-Whitney U test for non-normally distributed continuous data and Pearson chi-square test or Fisher’s exact test for categorical variables. Candidate predictors for DKD risk were evaluated and selected through univariate analysis. Multivariate logistic regression analysis was performed using backwards stepwise logistic regression. The variable entry and exit criteria were $p<$ 0.05 and $p>$ 0.10, respectively. The models for predicting the risk of DKD in DM patients were virtualized by the nomograms based on the independent variables extracted from multivariable logistic regression using the training set. In the nomogram, the impact of each variable on the endpoint event is reflected in their respective length and corresponding scores. Different patients have personalized scores, and the total scores related to each variable constitute the prediction probability.

The validation set was used to confirm the newly established nomograms. The concordance index (c-index) was used to evaluate the predictive accuracy of the models. Sensitivity and specificity were evaluated by receiver operating characteristics curve (ROC). Furthermore, the clinical value of the model was evaluated by decision curve analysis (DCA). The Delong test was used to compare the area under curve (AUC) of the ROC curves. A $p$ value $<$ 0.05 was considered as statistically significant.

3. Results

3.1 Demographic and clinical characteristics of patients

From March 2022 to March 2023, 597 patients (314 men and 283 women) with T2DM who met the inclusion criteria were retrospectively analyzed in this study. The mean age of the enrolled patients was (61.54 $\pm$ 11.80) years. 188 (31.49%) patients with T2DM were diagnosed as DKD by physicians. Among the 597 enrolled patients, 418 were assigned to the training set, while 179 were assigned to the validation set. The comparative clinical characteristics among the training and validation sets are shown in Table 1. The DKD rates between the training set and validation set were not significantly different ( $p=$ 0.944).

Table 1
The demographic data for the enrolled patients ( $n=$ 597)

Characteristics	Training set ( $n=$ 418)	Validation set ( $n=$ 179)	$p$ value
Gender, $n$ (%)			0.160
Male	212 (50.72)	102 (56.98)
Female	206 (49.28)	77 (43.02)
Age, years, mean $\pm$ SD	61.95 $\pm$ 11.57	60.58 $\pm$ 12.31	0.194
BMI, kg/m², mean $\pm$ SD	25.53 $\pm$ 3.53	25.56 $\pm$ 4.19	0.926
FPG, mmol/L, mean $\pm$ SD	9.81 $\pm$ 3.73	9.81 $\pm$ 3.84	0.993
HbA1c, %, mean $\pm$ SD	9.23 $\pm$ 2.28	9.77 $\pm$ 2.87	0.106
Creatinine, $\mu$ mol/L, median (IQR)	61.96 (49.84, 76.67)	62.48 (51.33, 78.38)	0.515
BUN, mmol/L, median (IQR)	5.55 (4.64, 6.97)	5.56 (4.62, 7.38)	0.615
Cys-C, mg/L, median (IQR)	0.90 (0.74, 1.17)	0.91 (0.78, 1.17)	0.232
TG, mmol/L, median (IQR)	1.50 (1.08, 2.33)	1.41 (1.03, 2.16)	0.138
TC, mmol/L, median (IQR)	4.59 (3.70, 5.43)	4.28 (3.67, 5.24)	0.218
HDL-C, mmol/L, median (IQR)	0.99 (0.81, 1.21)	1.00 (0.84, 1.18)	0.614
LDL-C, mmol/L, median (IQR)	2.70 (2.10, 3.30)	2.56 (2.05, 3.14)	0.115
Hypertension, $n$ (%)			0.785
Yes	266 (63.64)	116 (64.80)
No	152 (36.36)	63 (35.20)
DPN, $n$ (%)			0.274
Yes	251 (60.05)	116 (64.80)
No	167 (39.95)	63 (35.20)
DR, $n$ (%)			0.368
Yes	103 (24.64)	38 (21.23)
No	315 (75.36)	141 (78.77)
Diabetic foot, $n$ (%)			0.929
Yes	18 (4.31)	8 (4.47)
No	400 (95.69)	171 (95.53)
DD, $n$ (%)			0.430
$<$ 5 years	182 (43.54)	82 (45.81)
5–15 years	198 (47.37)	76 (42.46)
$>$ 15 years	38 (9.09)	21 (11.73)
DKD, $n$ (%)			0.944
Yes	132 (31.58)	56 (31.28)
No	286 (68.42)	123 (68.72)

Note: BMI, body mass index; SD, standard deviation; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin A1c; IQR, interquartile range; BUN, blood urea nitrogen; Cys-C, cystatin C; TG, triglyceride; TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; DPN, diabetic peripheral neuropathy; DR, diabetic retinopathy; DD, duration of diabetes; DKD, diabetic kidney disease.

3.2 Development of predictive models

The results of the univariate analysis of logistic regression model are presented in Table 2. We found that gender ( $p=$ 0.002), HbA1c levels ( $p<$ 0.001), Cys-C levels ( $p=$ 0.006), TC levels ( $p=$ 0.029), HDL-C levels ( $p$ = 0.004), presence of DPN ( $p=$ 0.028), presence of DR ( $p<$ 0.001) and DD were associated with the risk of DKD in patients with T2DM. All candidate predictors were considered in assessing the full model (model 1). After the multivariate analysis was done (Table 3), HbA1c levels (odds ratio [OR] $=$ 3.168, 95% confidence interval [CI] $=$ 2.488–4.034, $p<$ 0.001), HDL-C levels (OR $=$ 0.138, 95% CI $=$ 0.036–0.532, $p=$ 0.004), presence of DR (OR $=$ 2.721, 95% CI $=$ 1.250–5.925, $p=$ 0.012) and DD (5–15 years vs. $<$ 5 years, OR $=$ 10.024, 95% CI $=$ 4.469–22.483, $p<$ 0.001; $>$ 15 years vs. $<$ 5 years, OR $=$ 50.799, 95% CI $=$ 11.332–227.723, $p<$ 0.001) were identified as the predictors for DKD in model 1. Then, a model that incorporated the above independent predictors was developed and presented as a nomogram (Fig. 1A). We also developed a laboratory-based model (model 2) using the laboratory indexes. After the multivariate analysis was done (Table 4), HbA1c levels (OR $=$ 3.060, 95% CI $=$ 2.458–3.808, $p<$ 0.001), Cys-C levels (OR $=$ 2.337, 95% CI $=$ 1.485–3.679, $p<$ 0.001) and HDL-C levels (OR $=$ 0.241, 95% CI $=$ 0.073–0.798, $p=$ 0.020) were identified as the predictors for DKD in model 2. A nomogram constructed based on laboratory indexes was presented in Fig. 1B.

Table 2
Univariate analysis of predictors for diabetic kidney disease screening in the training cohort

Characteristics	$\beta$	OR (95% CI)	$p$ value
Gender
Male	Reference	Reference
Female	$-$ 0.677	0.508 (0.333, 0.775)	0.002
Age, years	0.009	1.009 (0.991, 1.028)	0.311
BMI, kg/m²	0.020	1.020 (0.962, 1.082)	0.502
FPG, mmol/L	0.036	1.037 (0.982, 1.095)	0.191
HbA1c, %	1.072	2.922 (2.377, 3.592)	$<$ 0.001
Creatinine, $\mu$ mol/L	0.001	1.001 (0.999, 1.004)	0.183
BUN, mmol/L	0.001	1.000 (0.998, 1.001)	0.657
Cys-C, mg/L	0.434	1.544 (1.534, 2.102)	0.006
TG, mmol/L	0.093	1.098 (0.991, 1.217)	0.075
TC, mmol/L	$-$ 0.191	0.826 (0.696, 0.981)	0.029
HDL-C, mmol/L	$-$ 1.116	0.327 (0.153, 0.702)	0.004
LDL-C, mmol/L	$-$ 0.162	0.850 (0.664, 1.088)	0.196
Hypertension
No	Reference	Reference
Yes	0.193	1.213 (0.786, 1.872)	0.382
DPN
No	Reference	Reference
Yes	$-$ 0.469	0.626 (0.412, 0.951)	0.028
DR
No	Reference	Reference
Yes	0.986	2.681 (1.690, 4.251)	$<$ 0.001
Diabetic foot
No	Reference	Reference
Yes	0.084	1.087 (0.399, 2.963)	0.870
DD
$<$ 5 years	Reference	Reference
5–15 years	1.449	4.258 (2.560, 7.080)	$<$ 0.001
$>$ 15 years	2.735	15.415 (6.801, 34.940)	$<$ 0.001

Note: OR, odds ratio; CI, confidence interval; Note: BMI, body mass index; SD, standard deviation; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin A1c; BUN, blood urea nitrogen; Cys-C, cystatin C; TG, triglyceride; TC, total cholesterol; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; DPN, diabetic peripheral neuropathy; DR, diabetic retinopathy; DD, duration of diabetes.

Table 3

Multivariate analysis of predictors for diabetic kidney disease screening in model 1

Characteristics	$\beta$	OR (95% CI)		$p$ value
HbA1c, %	1.153	3.168	(2.488, 4.034)	$<$ 0.001
HDL-C, mmol/L	$-$ 1.980	0.138	(0.036, 0.532)	0.004
DR
No	Reference	Reference
Yes	1.001	2.721	(1.250, 5.925)	0.012
DD
$<$ 5 years	Reference	Reference
5–15 years	2.305	10.024	(4.469, 22.483)	$<$ 0.001
$>$ 15 years	3.928	50.799	(11.332, 227.723)	$<$ 0.001

Note: OR, odds ratio; CI, confidence interval; HbA1c, glycated hemoglobin A1c; HDL-C, high density lipoprotein cholesterol; DR, diabetic retinopathy; DD, duration of diabetes.

Table 4

Multivariate analysis of predictors for diabetic kidney disease screening in model 2

Characteristics	$\beta$	OR (95% CI)	$p$ value
HbA1c, %	1.118	3.060 (2.458, 3.808)	$<$ 0.001
Cys-C, mmol/L	0.849	2.337 (1.485, 3.679)	$<$ 0.001
HDL-C, mmol/L	$-$ 1.422	0.241 (0.073, 0.798)	0.020

Note: OR, odds ratio; CI, confidence interval; HbA1c, glycated hemoglobin A1c; Cys-C, cystatin C; HDL-C, high density lipoprotein cholesterol.

Figure 1.

The nomograms for estimating risk of DKD in patients with T2DM: (A) full model (model 1); (B) laboratory-based model (model 2). HbA1c, glycated hemoglobin A1c; Cys-C, cystatin C; HDL-C, high density lipoprotein cholesterol; DR, diabetic retinopathy; DD, duration of diabetes; DKD, diabetic kidney disease; T2DM, type 2 diabetes mellitus.

3.3 Validation of the models

Figure 2.

ROC curves of different models for diabetic kidney disease screening: (A) ROC curve of model 1 in training set; (B) ROC curve of model 1 in validation set; (C) ROC curve of model 2 in training set; (D) ROC curve of model 2 in validation set. ROC: Receiver operating characteristic; AUC: area under curve.

The C-index for model 1 was 0.947 (95% CI, 0.923–0.971) in the training set (Fig. 2A) and 0.897 (95% CI, 0.847–0.948, Fig. 2B) in the validation set. The C-index for model 2 was 0.925 (95% CI, 0.896–0.954) in the training set (Fig. 2C) and 0.867 (95% CI, 0.812–0.922, Fig. 2D) in the validation set. Both models generated AUCs with good discrimination. For both models, the Hosmer-Lemeshow test found no statistical significance in the validation set ( $p=$ 0.112 and $p=$ 0.424, respectively), indicating that no departure from a perfect fit was found. The calibration curves demonstrated a good agreement between prediction and observation in the validation set for model 1 and model 2 (Fig. 3). To evaluate the clinical usefulness of the model, we conducted DCA in the validation set. The DCA curve showed that both models achieved a net benefit across all threshold probabilities (Fig. 4).

Figure 3.

Calibration curves of different models in the validation set: (A) calibration curves of model 1; (B) calibration curves of model 2.

Figure 4.

Decision curve analysis for the predictive models. The net benefit was produced against the high-risk threshold.

Subsequently, we compared the diagnostic performances of the two models in the validation set. The results showed that the AUC of model 1 for predicting DKD was similar to that of model 2 (0.897 vs. 0.867, $p=$ 0.138, Fig. 5). These results indicated that both models had a good diagnostic value in predicting DKD The diagnostic performances of the two models were not significantly different.

Figure 5.

Comparison of ROC curves of the established models.

4. Discussion

We enrolled all T2DM patients regardless of the renal function. We successfully constructed two nomograms (a full model and a laboratory-based model) for DKD screening. We demonstrated that the two models showed good discrimination and calibration for estimating DKD risk.

Using the predictors included HbA1c levels, HDL-C levels, presence of DR and DD, we constructed a full model (model 1) for predicting DKD. And using the laboratory indexes of HbA1c, HDL-C, and Cys-C, we developed a laboratory-based model (model 2). HbA1c is traditionally used to monitor blood glucose control in DM patients. It has been reported that lowering HbA1c could reduce the risk of micro- and macrovascular complications of DM [11]. Maintenance of the low HbA1c level can prevent the development and progression of DKD [12]. In previous studies, HbA1c was reported to be a sensitive risk factor for clinical outcomes in DM patients with DKD [8, 13]. The HbA1c-based predicting models were found to be helpful for differential diagnosis of DKD and non-DKD in DM patients [5, 14]. In line with previous studies, we found that higher HbA1c levels were associated with higher risk of DKD in the present study.

HDL-C concentration is one of the main features of dyslipidemia in patients with DM [15]. Several studies have demonstrated that high HDL-C level is the protective factor against the progression of DKD [15, 16]. In Italy, Russo et al. analyzed the data from 47,177 patients with T2DM and found that low HDL-C was an independent risk factor for the development of DKD [17]. In this study, we also identified that the HDL-C level was a predictor of DKD risk in T2DM patients, which was included in the two models we established. Our results showed that patients with a higher HDL-C level had a lower risk to develop DKD.

DR and DKD are two most important complications of DM. Previous studies have shown that there is a strong correlation between the severity of DR with severity of DKD [18]. Consistent with these studies, the presence of DR was also demonstrated as a predictor for DKD risk in T2DM patients in model 1. We found that patients with DR were more likely to develop DKD.

It is generally believed that the interval between T2DM onset and development of DKD was over 10 years [5]. Thus, a longer duration of T2DM was significantly associated with a higher risk of DKD [19, 20]. In the study where a specific time period was clearly defined, a duration of less than 5 years from T2DM onset could predict non-DKD [20]. Similarly, in this study, we found that patients with a DD more than 5 years had a higher risk to develop DKD than those with a DD less than 5 years. DD was identified as a predictor for DKD risk and included in our full model.

In our study, we also observed that a higher Cys-C level could predict a risk of DKD. It has been clearly pointed out that serum Cys-C is an early marker for detection of DKD and reflection of renal function changes [21, 22]. Wei et al. [23] found in their study that patients with DKD had a higher Cys-C level than those without DKD. Liao et al. [24] conducted a meta-analysis to evaluate the accuracy of serum Cys-C for diagnosing DKD. They included 26 published studies and found that Cys-C had an excellent diagnostic value with good sensitivity and specificity for DKD [24]. Thus, Cys-C is potential risk factor for DKD development in patients with T2DM. In the present study, Cys-C level was included in our laboratory-based model. But for the full model, Cys-C level was rejected after multivariate analysis. The rejection of this factor in model 1 may be the result of nuances in the data set or confounding by other predictors.

The risk screening tools are essential and cost-effective in clinical practice and healthcare management. In this study, we demonstrated that both models we constructed exhibited good discrimination and calibration in our internal validations. They are the potential useful tools for DKD screening in patients with T2DM. Furthermore, the diagnostic performances of the two models were not significantly different. Physicians can choose different models to evaluate DKD risk for patients with T2DM based on the personal medical data.

The present study has several limitations. First, the nomograms were established based on data obtained from a single center in China. This may cause selection bias. Second, information bias and selection bias (including inclusion criteria and processing of missing data) may cause bias in the results of this study. Third, DKD is usually associated with numerous factors. But our study only included the most available factors. In this study, the influencing factors of DKD were not fully included. Artificial intelligence has broad prospects in disease detection [25, 26]. In the future, we will include more influencing factors [27] and expand the sample size for further research. In the future, the prediction model established in this study will be combined with artificial intelligence for early diagnosis and classification of domain diseases [28].

5. Conclusion

In this study, we successfully constructed two prediction models to evaluate the risk of DKD in patients with T2DM. The two models exhibited good predictive performance and could be recommended for DKD screening and early detection.

Ethics statement

This retrospective study was approved by the Ethics Committee of Bengbu First People’s Hospital in accordance with the Declaration of Helsinki. Due to the nature of retrospective study and anonymized patient’s information, informed consent was waived.

Availability of data and materials

All data generated or analyzed during this study are included in this published article.

Funding

The authors did not receive any financial support for the research, authorship, or publication of this manuscript.

Author contributions

Conception and design of the work: CX; Data collection: CX, PXM; Analysis and interpretation of the data: CX, PXM; Statistical analysis: CX, PXM; Drafting the manuscript: CX; Critical revision of the manuscript: CX, PXM. All authors approved the final version of the manuscript.

Footnotes

Acknowledgments

Not applicable.

Conflict of interest

None of the authors have any personal, financial, commercial, or academic conflicts of interest to report.

References

Lin

Chang

Yang

, et al. Update of pathophysiology and management of diabetic kidney disease. J Formos Med Assoc. 2018 Aug; 117(8): 662-675. doi: 10.1016/j.jfma.2018.02.007.

Tuttle

Agarwal

Alpers

, et al. Molecular mechanisms and therapeutic targets for diabetic kidney disease. Kidney Int. 2022 Aug; 102(2): 248-260. doi: 10.1016/j.kint.2022.05.012.

Tye

Denig

Heerspink

HJL

. Precision medicine approaches for diabetic kidney disease: Opportunities and challenges. Nephrol Dial Transplant. 2021 Jun 22; 36(Suppl 2): 3-9. doi: 10.1093/ndt/gfab045.

Hung

Hsu

Chen

, et al. Recent advances in diabetic kidney diseases: From kidney injury to kidney fibrosis. Int J Mol Sci. 2021 Nov 1; 22(21): 11857. doi: 10.3390/ijms222111857.

Tan

Choo

JCJ

Fook-Chong

, et al. Development and validation of a novel nomogram to predict diabetic kidney disease in patients with type 2 diabetic mellitus and proteinuric kidney disease. Int Urol Nephrol. 2023 Jan; 55(1): 191-200. doi: 10.1007/s11255-022-03299-x.

Fiorentino

Bolignano

Tesar

, et al. Renal biopsy in patients with diabetes: A pooled meta-analysis of 48 studies. Nephrol Dial Transplant. 2017 Jan 1; 32(1): 97-110. doi: 10.1093/ndt/gfw070.

Slieker

van der Heijden

AAWA

Siddiqui

Langendoen-Gort

Nijpels

Herings

Feenstra

Moons

KGM

Bell

Elders

’t Hart

Beulens

JWJ

. Performance of prediction models for nephropathy in people with type 2 diabetes: Systematic review and external validation study. BMJ. 2021 Sep 28; 374: n2134. doi: 10.1136/bmj.n2134. PMID: 34583929; PMCID: PMC8477272.

Jiang

Wang

Shen

Wang

Gao

Liu

Zheng

Chang

Yang

Chang

. Establishment and validation of a risk prediction model for early diabetic kidney disease based on a systematic review and meta-analysis of 20 cohorts. Diabetes Care. 2020 Apr; 43(4): 925-933. doi: 10.2337/dc19-1897. PMID: 32198286.

Grimes

. The nomogram epidemic: Resurgence of a medical relic. Ann Intern Med. 2008 Aug 19; 149(4): 273-5. doi: 10.7326/0003-4819-149-4-200808190-00010.

10.

Buse

Wexler

Tsapas

, et al. 2019 Update to: Management of Hyperglycemia in Type 2 Diabetes, 2018. A Consensus Report by the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetes Care. 2020 Feb; 43(2): 487-493. doi: 10.2337/dci19-0066. Epub 2019 Dec 19. Erratum in: Diabetes Care. 2020 Jul; 43(7): 1670.

11.

Yan

Kondo

Oniki

, et al. Predictive Ability of Visit-to-Visit Variability of HbA1c Measurements for the Development of Diabetic Kidney Disease: A Retrospective Longitudinal Observational Study. J Diabetes Res. 2022 Jan 22; 2022: 6934188. doi: 10.1155/2022/6934188.

12.

Nakanishi

Hirukawa

Shimoda

, et al. Comparison of HbA1c levels and body mass index for prevention of diabetic kidney disease: A retrospective longitudinal study using outpatient clinical data in Japanese patients with type 2 diabetes mellitus. Diabetes Res Clin Pract. 2019 Sep; 155: 107807. doi: 10.1016/j.diabres.2019.107807.

13.

Kuo

Lin

Niu

, et al. Glycated hemoglobin and outcomes in patients with advanced diabetic chronic kidney disease. Sci Rep. 2016 Jan 28; 6: 20028. doi: 10.1038/srep20028.

14.

Wang

Qiao

, et al. Screening tools based on nomogram for diabetic kidney diseases in chinese type 2 diabetes mellitus patients. Diabetes Metab J. 2021 Sep; 45(5): 708-718. doi: 10.4093/dmj.2020.0117.

15.

Yun

Kim

, et al. Risk factors for the development and progression of diabetic kidney disease in patients with type 2 diabetes mellitus and advanced diabetic retinopathy. Diabetes Metab J. 2016 Dec; 40(6): 473-481. doi: 10.4093/dmj.2016.40.6.473.

16.

Lee

Wang

Huang

, et al. High triglyceride-to-HDL cholesterol ratio associated with albuminuria in type 2 diabetic subjects. J Diabetes Complications. 2013 May-Jun; 27(3): 243-7. doi: 10.1016/j.jdiacomp.2012.11.004.

17.

Russo

De Cosmo

Viazzi

, et al. Plasma Triglycerides and HDL-C Levels Predict the Development of Diabetic Kidney Disease in Subjects With Type 2 Diabetes: The AMD Annals Initiative. Diabetes Care. 2016 Dec; 39(12): 2278-2287. doi: 10.2337/dc16-1246.

18.

Saini

Kochar

Poonia

. Clinical correlation of diabetic retinopathy with nephropathy and neuropathy. Indian J Ophthalmol. 2021 Nov; 69(11): 3364-3368. doi: 10.4103/ijo.IJO_1237_21.

19.

Kritmetapak

Anutrakulchai

Pongchaiyakul

, et al. Clinical and pathological characteristics of non-diabetic renal disease in type 2 diabetes patients. Clin Kidney J. 2018 Jun; 11(3): 342-347. doi: 10.1093/ckj/sfx111.

20.

Wang

Han

Zhao

, et al. Identification of clinical predictors of diabetic nephropathy and non-diabetic renal disease in Chinese patients with type 2 diabetes, with reference to disease course and outcome. Acta Diabetol. 2019 Aug; 56(8): 939-946. doi: 10.1007/s00592-019-01324-7.

21.

Elsayed

El Badawy

Ahmed

, et al. Serum cystatin C as an indicator for early detection of diabetic nephropathy in type 2 diabetes mellitus. Diabetes Metab Syndr. 2019 Jan-Feb; 13(1): 374-381. doi: 10.1016/j.dsx.2018.08.017.

22.

Wang

, et al. Diagnostic value of the combined measurement of serum hcy, serum cys C, and urinary microalbumin in type 2 diabetes mellitus with early complicating diabetic nephropathy. ISRN Endocrinol. 2013 Sep 18; 2013: 407452. doi: 10.1155/2013/407452.

23.

Wei

Wang

Shen

, et al. Diagnostic value of triglyceride and cystatin C ratio in diabetic kidney disease: A retrospective and prospective cohort study based on renal biopsy. BMC Nephrol. 2022 Jul 27; 23(1): 270. doi: 10.1186/s12882-022-02888-3.

24.

Liao

Zhu

Xue

. Diagnostic value of serum cystatin C for diabetic nephropathy: A meta-analysis. BMC Endocr Disord. 2022 Jun 2; 22(1): 149. doi: 10.1186/s12902-022-01052-0.

25.

Shahamiri

Thabtah

Abdelhamid

. A new classification system for autism based on machine learning of artificial intelligence. Technol Health Care. 2022; 30(3): 605-622. doi: 10.3233/THC-213032. PMID: 34657857.

26.

Choudhury

Perumalla

. Detecting breast cancer using artificial intelligence: Convolutional neural network. Technol Health Care. 2021; 29(1): 33-43. doi: 10.3233/THC-202226. PMID: 32444590.

27.

Zhang

Xin

. Association between serum magnesium and common complications of diabetes mellitus. Technol Health Care. 2018; 26(S1): 379-387. doi: 10.3233/THC-174702. PMID: 29758962; PMCID: PMC6004978.

28.

Preston

Meng

Burgess

Ferdousi

Azmi

Petropoulos

Kaye

Malik

Zheng

Alam

. Artificial intelligence utilising corneal confocal microscopy for the diagnosis of peripheral neuropathy in diabetes mellitus and prediabetes. Diabetologia. 2022 Mar; 65(3): 457-466. doi: 10.1007/s00125-021-05617-x. Epub 2021 Nov 21. PMID: 34806115; PMCID: PMC8803718.

Developing screening tools to estimate the risk of diabetic kidney disease in patients with type 2 diabetes mellitus

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Methods

2.1 Patients

2.2 Data collection

2.3 Predicted outcomes

2.4 Statistical methods

3. Results

3.1 Demographic and clinical characteristics of patients

Table 1 The demographic data for the enrolled patients ( n = 597)

Table 2 Univariate analysis of predictors for diabetic kidney disease screening in the training cohort

5. Conclusion

Ethics statement

Availability of data and materials

Funding

Author contributions

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
The demographic data for the enrolled patients ( $n=$ 597)

Table 2
Univariate analysis of predictors for diabetic kidney disease screening in the training cohort