In this study, we identified four subgroups of T2DM using cluster analysis based on five baseline variables, including HbA1c, BMI, age at diagnosis, HOMA2-β, and HOMA2-IR. These subgroups were identified as the severe insulin deficiency diabetes (SIDD), mild insulin resistance diabetes (MIRD), moderate glycemic control diabetes (MGCD), and moderate weight insulin deficiency diabetes (MWIDD) subgroups, respectively. Every subgroup had its own unique phenotypic and biochemical characteristics. Firstly, the rationale for choosing the parameter was based on whether it is the major factor affecting the course of T2DM and the risk of complications. For example, HbA1c is generally accepted as the gold standard for blood glucose control. HOMA2-β and HOMA2-IR are the standardized quantitative indicators of insulin secretion and insulin resistance, respectively. Age and BMI are regarded as the important clinical variables associated with the risk of diabetic complications. Secondly, the clinical utility of each parameter was also taken into account. The parameters that are usually monitored in routine clinical practice were included in our classification system. The subjects enrolled in our study had a comparatively long history of T2DM, which is different from previous study focusing on newly diagnosed patients with diabetes [15, 16, 23, 24]. Our data revealed that these groups had significant different risks of proteinuria progression and renal decline during the follow-up period.

Only the SIDD subgroup was in line with the cluster identified in the Scania cohort; the other three subgroups (MIRD, MGCD, and MWIDD) had their own unique phenotypic and biochemical characteristics. In our study, the two subgroups characterized by severe insulin deficiency (SIDD and MWIDD) accounted for 52.3% of the total population. The data showed that the incidence of SIDD (23.9%) was remarkably high in our cohort compared to European cohorts [15, 24] and Asian cohorts such as those from Japan and Thailand [16, 25], but it was similar to that of a Ukrainian cohort (25%) with a similar diabetes history [26]. First of all, this is probably due to the higher proportion of patients with T2DM in our cohort who had a longer diabetes history and worse insulin secretion function [27, 28]. Secondly, the East Asian population, including the Chinese population, displayed a lower insulin secretion capacity [29, 30] as compared with the white population. However, the proportion with SIDD in our cohort was lower than in a cohort of newly diagnosed patients with T2DM in India [31]. This discrepancy highlights the important roles of race, region, and genetic variation in determining the distribution of subgroups based on the novel classification of T2DM populations. Although the MWIDD subgroup was similar to the SIDD subgroup in terms of its reduced insulin secretion and poor glycemic control, the MWIDD subgroup had a lower BMI and insulin resistance. In addition, the metabolic-syndrome-related indicators in the MWIDD subgroup were significantly better than those in the SIDD subgroup. The MWIDD subgroup could be regarded as a new subgroup with unique clinical features that has not been reported in previous studies [15, 23, 32, 33]. The MIRD subgroup in our study was distinguished by a higher BMI and mild insulin resistance, but the BMI and HOMA2-IR in this subgroup were lower than those in SIRD subgroups from Sweden, Germany, and the United States [15, 24, 34]. A previous study indicated that patients with T2DM in China and India are typically diagnosed at a younger age and exhibit weaker insulin secretion, lower insulin resistance, and a lower BMI [35]. Additionally, it has been demonstrated that insulin resistance is positively correlated with age in Americans but negatively correlated with age in Chinese [36]. The MGCD subgroup was characterized by a low HbA1c level and moderate levels of HOMA2-IR, BMI, and HOMA2-β. The clinical phenotype of this subgroup was similar to the MARD (mild age-related diabetes) subgroup described in European cohorts and the milder MD (mild diabetes) subgroup identified by Slieker [15, 23, 24, 32].

After adjustment of the known risk factors, both the MIRD and SIDD subgroups were identified as independent risk factors for DKD in our T2DM cohort. It was demonstrated that the MIRD group exhibited an increased risk of progression from non-albuminuria to micro-albuminuria. Previous clinical studies have shown that a decrease in insulin sensitivity is significantly related to the production of proteinuria [37,38,39]. It has been reported that insulin resistance is relevant to renal hemodynamic changes. Hyperinsulinemia may increase glomerular hydrostatic pressure, increase renal vascular permeability, aggravate glomerular hyperfiltration, and enhance renal sodium reabsorption. In conclusion, the cascade reaction caused by insulin resistance may change the homeostasis of renal endothelial function and the coordination of hemodynamics [40,41,42]. The remarkable difference is that the insulin resistance in our study was mild, which is inconsistent with the previously clustered subgroups in northern Europe, Asia, and Ukraine, which showed severe insulin resistance [23, 26, 43]. Although the other risk factors were adjusted, the MIRD subgroup was still an independent predictor of the onset of proteinuria in patients with T2DM in our cohort. Therefore, insulin resistance in patients with T2DM should be continuously managed, especially for patients who have not yet developed proteinuria. Drugs that are known to improve insulin resistance and body weight, such as metformin and SGLT-2 inhibitors, have been shown to reduce urinary protein excretion and present increased heart and kidney benefits [44, 45].

Furthermore, the SIDD group had the highest risk of progression from micro-albuminuria to macro-albuminuria in our cohort. This finding suggests that the progression of proteinuria in DKD is likely to be more affected by insulin deficiency and poor glycemic control. This indicates that insulin deficiency and hyperglycemia could promote inflammation and oxidative stress and damage podocyte function [46, 47]. Podocytes play a pivotal role in maintaining glomerular structure and function, and damage to podocytes will further aggravate the degree of proteinuria. It needs to be mentioned that the SIDD subgroup showed high metabolic disorder, with high BMI, blood glucose, blood uric acid, and blood lipid levels, all of which are considered to be independent risk factors for DKD progression. Previous studies indicated that the risk of DKD was higher in the SIRD subgroup, which is characterized by severe insulin resistance [16, 19, 23, 24]. Meanwhile, there are reports from India which show that the risk of DKD is higher in the SIDD subgroup (with severe insulin deficiency) and the CIRDD (combined insulin resistant and deficient diabetes) subgroup (with both severe insulin resistance and insulin deficiency) [18, 31]. Consistent with the studies mentioned above, our study also suggests that patients with T2DM characterized by insulin resistance or insulin deficiency have a higher risk of the onset or progression of DKD.

Unexpectedly, no statistically significant differences in renal decline events between these four subgroups were found during the investigation. This was probably due to the limitation caused by the short observation period, as there are few rapid renal-function decline events in a short period. Moreover, most of the included T2DM population was in the early and middle stages of DKD according to the eGFR.

The clustering based on the five clinical variables performed for patients with T2DM with a longer disease history in our study may be different from the previously reported clustering of a newly diagnosed population [15, 24, 48]. The inclusion of patients with different disease management strategies may have yielded potential confounding factors. Moreover, information on the drug interventions of enrolled patients with T2DM was not collected in our analysis, which might have led us to ignore the complex interaction between drug therapy and DKD progression. In addition, traditional scoring systems such as the Kidney Failure Risk Equation (KFRE) were not used to assess the association of clustering-based subgroups and DKD progression. Besides this, information on the family history of CKD, albuminuria, and the presence of hypertension in each included patient was not collected in our analysis, which might have led us to ignore the complex role of genetic susceptibility in predicting DKD progression. Therefore, we need to further verify the stability and applicability of the novel subgroup classification, and our findings need to be validated in future muti-center prospective studies.