Diabetes mellitus (DM) is a common clinical chronic disease with high rate of morbidity, long duration, and lifelong elevation of glucose levels, accompanied by insufficient secretion of insulin by pancreatic islet β-cells or insufficient biological effect of insulin in the peripheral tissues (1). According to the International Diabetes Federation (IDF), by 2040, there will be 642 million diabetics globally, up from the current projection of 460 million (2). Thus, diabetes has now become a serious and urgent public health problem which requires prior attention.

DM can be classified into type 1 diabetes mellitus (T1DM), type 2 diabetes mellitus (T2DM), other special types of diabetes mellitus and gestational diabetes mellitus on the basis of etiological characteristics. The majority of DM cases are classified as T2DM, accounting for 90-95% of the total diabetic patients, and occur as a result of insulin resistance combined with insufficient insulin secretion. T1DM is the second most common, accounting for 5-10% of the diabetic patients (3). T1DM is caused by insulin insufficiency due to pancreatic β-cells dysfunction or autoimmune disruption, and the patient is dependent on insulin for life (4, 5). Approximately, 5-15% of the adults diagnosed with T2DM may actually have T1DM (6). As a result, up to 50% of the actual T1DM cases may be misdiagnosed as T2DM, i.e., the overall number of T1DM cases is considerably underestimated (7). An accurate differential diagnosis is, therefore, essential for the optimal treatment and avoidance of complications.

Distinction between T1DM and T2DM can usually be done according to clinical criteria, primarily based on clinical presentation, age, and body mass index (BMI). However, owing to disease complexity and the diversity of the population, it is difficult to diagnose the phenotype. Regardingly, the laboratory tests can help to differentiate between T1DM and T2DM by using metabolic tests such as C-peptide, insulin, and insulinogen. In a study by Bolinder (8), total insulinogen and intermediate insulinogen degradation products were measured in the subjects, and there was considerable overlap in the levels of total and intermediate insulinogen between T1DM and T2DM patients 3-4 months after the onset of DM. Katz et al. (9) also showed that C-peptide levels in T1DM and T2DM patients were overlapped. Therefore, the use of these laboratory tests in the differential diagnosis of T1DM and T2DM is constrained.

T1DM is caused by autoimmune β-cells destruction, whereas T2DM is caused by insulin resistance, which causes relative β-cells failure, eventually (10). Therefore, if autoantibodies targeting β-cells are detected, this is suggestive of an autoimmune etiology and can help to diagnose autoimmune T1DM (11). There are currently five diabetes autoantibodies that can be used to diagnose T1DM and predict disease progression in non-T1DM patients, which include islet cells autoantibodies (ICA), insulin autoantibodies (IAA), glutamic acid decarboxylase autoantibodies (GADA), islets antigen 2 autoantibodies (IA-2A), and zinc transporter-8 autoantibodies (ZnT8A) (12). Of these, GADA, ICA and IAA are considered to be the three most important antibodies (13). About 70–80% of the individuals with newly diagnosed T1DM have ICA and GADA. 60% have IA-2A and ZnT8A identified, while IAA is infrequent in adults with newly diagnosed T1DM but is present in 50–60% of adolescents with the disease (14). As the duration of diabetes increases, fewer people remain positive for diabetes autoantibodies other than anti-insulin antibodies (14). As in the case of ICA, they are present for a shorter period of time, appearing only in the early stages of T1DM (15). Moreover, the use of insulin is associated with the production of insulin antibodies, and it is impossible to distinguish between insulin antibodies and IAA after 14 days of insulin treatment (16). In addition, overweight or obese adults with a clinical diagnosis of T2DM may also present with positive diabetic autoantibodies (17). In summary, the diabetic autoantibody profile has limitations in differentiating between T1DM and T2DM.

Some studies have shown that cell counts (18–20), liver function (21, 22), blood lipids (23) levels, etc. are linked with the incidence of DM. To investigate the value of diabetic autoantibodies in combination with the clinical data and routine laboratory indicators in the classification of DM, the present study included 522 diabetic patients. The factors of interest including gender, age, autoantibody profile, C-peptide levels, and other relevant variables were recorded. Logistic regression and nomogram analysis were employed to develop predictive models for T1DM, with the goal of improving risk stratification and guiding personalized interventions for individuals at risk of developing diabetes.

Diabetes mellitus (DM) is a group of metabolic diseases characterized by hyperglycemia caused by defects in either insulin secretion or action, or both. Diabetes-related chronic hyperglycemia is linked to long-term damage, malfunction, and failure of several organs, including the heart, blood vessels, kidneys, eyes, and nerves (24). The healthcare expenditures and access to treatment are unequal between developed and developing countries, nevertheless, both bear a huge financial burden (25, 26). Early and accurate identification of DM is important for determining treatment options, improving outcomes and reducing the economic burden. However, there are limitations in classifying T1DM and T2DM based on the clinical data, laboratory metabolic testing, and diabetes autoantibody testing (8, 9, 14–17). Additionally, the value of routine laboratory tests for typing is unclear. In this study, we constructed a quantifiable and simple nomogram for predicting T1DM, which can help clinicians to differentiate between T1DM and T2DM, and which contains one clinical parameter (age), two routine laboratory tests (PA, HDL-C), one islet β-cells function assessment test (C-peptide) and three diabetic autoantibody tests (GADA, ICA, IA-2A).

There is a clear distinction between the age of onset of T1DM and T2DM. The onset of T1DM is usually between 5-7 years of age and adolescence, but can occur at any age (7), while the onset of T2DM occurs after puberty. In the present study, the mean age of patients in the T1DM group was 28.7 years, which was significantly lower than that of the T2DM group, which was 58.5 years (p < 0.05). Moreover, in the Logistic regression model, the regression coefficient of age was 0.012 (p < 0.05), indicating that age is a significant influence on DM typing.

Prealbumin (PA) is a negative acute phase response protein and non-specific host defense substance, mainly synthesized by the liver, with a half-life of about 2 days, which makes it more sensitive compared to albumin, which has a half-life of 20 days, and has been used in clinical practice mainly to assess hepatic impairment and malnutrition (27). In recent years, it has been discovered that PA contributes to autoimmune diseases. One study (28) has shown that PA levels were negatively correlated with the degree of autoimmunity, which is consistent with the negative correlation between PA and T1DM typing in this study. In the present study, PA levels in the T2DM group were significantly higher than those in the T1DM group and were positively correlated with T2DM, which confirms that patients with T2DM are more prone to cardiovascular disease (29). Nicoletta Dozio et al. (30) also showed that PA levels vary at different stages of T1DM disease course, with lower levels in patients with longer disease duration, and this study is expected to play a role in evaluating the stage of disease and prognosis of patients at the time of initial diagnosis of T1DM.

High-density lipoprotein cholesterol (HDL-C) has anti-atherosclerotic and antioxidant properties and prevents oxidized LDL (ox-LDL) interceded endothelial dysfunction (31). Characteristic dyslipidemia usually precedes the diagnosis of T2DM, such as reduced HDL-C levels, suggesting that reduced HDL-C promotes the onset and progression of T2DM and diabetic vascular complications (32). Indeed, it has been found that there is a bidirectional association between HDL-C and T2DM, whereby hyperglycemia and hyperinsulinemia occurring in T2DM may lead to reduced HDL-C levels and deterioration of HDL function through various alterations in the HDL particles proteome and lipidome (33).Thus, via altering insulin secretion, peripheral insulin sensitivity, non-insulin-dependent glucose uptake, and adipose tissue metabolic activation, HDL-C may also have an impact on glucose homeostasis (34). In the present study, the mean HDL-C values in the T2DM group were lower than those in the T1DM group, and there was a negative correlation between the HDL-C values and the T2DM phenotype, which is consistent with the findings mentioned above.

As described in 1967 (35), C-peptide is a 31-amino acid peptide, facilitating the correct folding of insulin and formation of its disulfide bridges. Proinsulin is cleaved into insulin and C-peptide. These two proteins are stored in the secretory granules of the pancreatic β-cells and eventually released together in equimolar amounts. C-peptide has negligible extraction by the liver and constant peripheral clearance. Its half-life is longer than insulin (20–30 vs. 3–5 min) (36). Therefore, the physiology of C-peptide makes it appropriate for assessing insulin secretion. Absolute insulin deficiency is a key feature of Type 1 diabetes, and C-peptide levels taken within the first few years of diagnosis may be useful in confirming Type 1 diabetes if results are low (37). As such, C-peptide has been a valuable tool in elucidating the pathophysiology of T1DM and T2DM. In the present study, the mean C-peptide values in the T2DM group were higher than those in the T1DM group, which is consistent with the above-mentioned findings.

Autoimmunity and cellular immunity in T1DM patients contribute to the onset and progression of the disease. Pertinently, some scholars have proposed that diabetic autoantibody detection in patients’ serum can be an effective diagnostic method for typing of diabetic patients, and currently the clinical use of antibodies including GADA, IAA, IA-2A, ZnT8A, ICA (12). Among diabetic autoantibodies, the highest positive rate belongs to the Glutamic acid decarboxylase (GAD) antibody. GAD is a key enzyme in the synthesis of inhibitory neurotransmitter γ-aminobutyric acid, and the available data confirm that its level can be elevated several years or even more than 10 years prior to the onset of T1DM. Moreover, it has the characteristics of high sensitivity and specificity, and is considered to be a specific marker for immune destruction of pancreatic islet β-cells in T1DM patients (38). GADA is the earliest antibody to GAD, and some scholars have found that a single positive GADA has a predictive value for insulin β-cell function (39). IAA was discovered in 1983 in T1DM patients who had not used exogenous insulin (40). Subsequent studies have shown that anti-insulin antibodies are present prior to the onset of T1DM (41), and were negatively correlated with age at onset of T1DM (42). ICA is a cytoplasmic antibody to pancreatic islet β-cells, which can cause an immune response upon binding to islet cell surface antigen, resulting in cytotoxic effects on islet cell cytoplasmic components, leading to cell lysis, death, and ultimately DM. Also, ICA is the first diabetic autoantibody found to be associated with the development of T1DM disease (43). According to earlier research, ICA is present in approximately 70% of T1DM patients, but for a short period of time i.e., appearing only in the early stages of T1DM (15). Positive ICA is now considered to be indicative of autoimmune damage to pancreatic β-cells and is highly predictive of T1DM when it is persistently positive or at high levels.

The results of the present study demonstrated that the difference in the positive rates of ZnT8A, ICA, IA-2A and GADA between patients in the T1DM and the T2DM group was statistically significant (p < 0.05), suggesting that the pancreatic β-cells had undergone a strong autoimmune reaction, which had caused impaired insulin secretion from the β-cells, leading towards pancreatic β-cell failure, which was in line with the main characteristics of T1DM. Between the two groups, there was no statistically significant difference in the positive rates of IAA (p > 0.05).

In this study, the following seven predictors were selected: age, PA, HDL-C, ICA, IA-2A, GADA, and C-peptide. A multivariate predictive model, nomogram, was established with excellent efficacy, and it could distinguish T1DM well, with an AUC of 0.966 and 0.961 in the training set and validation set, respectively. According to the calibration curves, the nomogram has a strong calibration. Moreover, it can serve as a useful tool for clinical applications and lower the cost and burden of disease, according to subsequent DCA and CIC assessments. Finally, we have built a web-based computational tool that may facilitate doctors’ by providing a platform for convenient and enhanced application of the nomogram. A previous study (44) built a similar predictive model that included age, body mass index, FPG, and TC to focus on the risk of developing T2DM in hypertensive patients. Another study (45) developed and validated a personalized prediction nomogram for non-obese adults with 5-year T2DM risk, including age, GGT, TG, FPG, HbA1c, and fatty liver. In our study, we introduced a novel predictive model integrating autoantibody profiles with clinical and laboratory data, and the differential prediction of T1DM and T2DM was carried out, which is a supplement to the former and the field. Unlike existing models that often rely on single diagnostic criteria or limited parameters, our approach aims to significantly enhance classification accuracy by considering a comprehensive set of predictors.

However, there are still some limitations in this study. First, our participants were all patients from the same hospital, which may make the results not applicable to other countries and regions. Second, we excluded patients with incomplete data, leading to potential selection bias inherent in our participant recruitment process. Future research efforts should prioritize addressing these limitations through larger, multicenter studies involving diverse patient populations and rigorous validation protocols. Such endeavors would enhance the generalizability and reliability of predictive models for classifying diabetes mellitus types. By expanding the scope of research beyond single-center studies and incorporating broader patient demographics, we can strengthen the evidence base supporting clinical decision-making in diabetes care.

Despite these limitations, our findings remain a supporting tool for clinical decision making. Our study demonstrates that the predictive nomogram integrating specific autoantibodies and laboratory indices can significantly improve the accuracy of diabetes mellitus classification. It provides quantitative predictions based on individual patient data, aiding clinicians in making informed decisions about treatment, monitoring requirements, and patient education efforts. However, implementing the nomogram in clinical settings may present challenges. Healthcare providers must ensure access to the necessary laboratory tests and interpret the nomogram’s results within the context of each patient’s clinical presentation and medical history. Training and education for healthcare providers on the use and interpretation of the nomogram are essential to optimize its utility and minimize the risk of misclassification or misinterpretation of results.

In conclusion, the application of individual clinical data, routine laboratory indicators or diabetes autoantibodies in the diagnosis and treatment of DM is relatively limited, and it is necessary to comprehensively consider age, PA, HDL-C, ICA, IA-2A, GADA, and C-peptide. Conclusively, the nomogram that is created based on these variables may provide useful differentiation between T1DM and T2DM, and the assessment of changes through the course of DM, which can provide a scientific guide to clinicians for diabetes prevention and treatment.