The present study successfully validated the performance of the MODY calculator in differentiating a large sample of patients with a confirmed diagnosis of MODY from patients with T1DM and T2DM and also succeeded in establishing a cut-off point for the probability of this diagnosis.
The originality of the present study lies in the application and validation of the MODY probability calculator in a large cohort of patients diagnosed with diabetes at an early age and who, furthermore, were not exclusively Caucasian but were of mixed race and of different ethnic origin, characteristics that make the Brazilian people a unique population worldwide [5, 6]. This study will allow the MODY calculator to be used within the country as a tool for estimating the probability of MODY prior to testing, thus screening and identifying individuals to be referred for genetic testing. This is particularly relevant since the frequency of MODY in the Brazilian population is uncertain and the diagnosis of this DM subtype is challenging [1].
Few studies involving patients with MODY have included non-Caucasian populations. Of note, Misra et al. [11] analyzed a population referred for genetic testing for MODY and compared the characteristics of the individuals of South Asian descent with those of Caucasians. The characteristics of South Asians with a diagnosis of MODY were found to be largely similar to those of the Caucasian patients with MODY, except for the fact that the South Asians had lower BMI and were younger at the time of diagnosis. The application of the MODY calculator could therefore be discriminatory in this population; however, the tool has yet to be validated for use in that ethnic group.
Non-Caucasian patients were also evaluated in a study conducted in Australia [12]. The prevalence of MODY and permanent neonatal diabetes mellitus was investigated in 1,668 European and non-European patients with a diagnosis of T1DM or T2DM. Of those patients, 196 were under 35 years of age, 148 were of European descent, and, according to the probability calculator, 8% had a high probability of MODY. After genetic testing in these individuals, a diagnosis of MODY was confirmed in 3 patients. However, in the non-European group of patients (n = 46), although 28% had a high probability of MODY according to the calculator, none had the diagnosis confirmed at genetic testing. For the patients clinically diagnosed with T1DM or T2DM, the cut-off value used for referring patients for genetic testing for MODY was 25% with the use of the calculator. The prevalence of MODY in the participants of European descent was 0.28% (95%CI: 0.09–0.77%) compared to 0 in the participants of non-European descent. This lower prevalence in that population could be due to the small number of individuals in that study, which could have limited the evaluation of the calculator. Furthermore, the cut-off point used for genetic testing was low, since it was based on the study conducted by Shields et al. [4], which suggests a cut off limit of between 10 and 25%.
In the present study, the principal factors distinguishing the MODY group from the T1DM and T2DM groups were BMI, a positive family history of diabetes and mean HbA1c value, with age at diagnosis only being a distinguishing factor when the MODY group was compared to the T2DM group.
The majority (85%) of the young people with T2DM had a BMI > 25 kg/m2 and this was an important distinguishing factor in the MODY patients, the majority (75%) of whom had a BMI within the normal range. Although approximately 35% of the patients in the present study with a diagnosis of T1DM were also overweight, there were few cases of obesity compared to the T2DM group.
In relation to family history, the finding that in 18.43% of MODY cases there was no family history of diabetes is intriguing. Our principal hypothesis for this finding is the absence of confirmed genetic testing in the parents or, in a minority of cases, the presence of a de novo mutation, as reported by Shields et al. [4] and in our recent description of a MODY cohort, which showed 24% of de novo mutations in GCK-MODY patients [13].
Most of the MODY patients were not in use of oral anti-diabetic drugs. This could be explained by the fact that this cohort consists primarily of individuals with GCK-MODY (70%) who typically show non-progressive mild hyperglycemia with no need for pharmacological treatment [2, 3, 13]. This characteristic of the present cohort was also reflected in the median HbA1c level of the MODY patients, which was significantly lower compared to that of the T1DM and T2DM groups.
Age at diagnosis was similar in the MODY group and the T1DM group. Therefore, as suggested by Shields et al. [4], completing the diagnostic investigation by measuring C-peptide and the main islet autoantibodies can be useful in distinguishing between MODY and T1DM, particularly in patients with an early diagnosis, who are thin and have been in use of insulin therapy since diagnosis. A typical individual with T1DM could be diagnosed at a time when pancreatic beta-cell reserve is not yet depleted, and that individual may present with measurable C-peptide levels over the first five years of the disease. Likewise, positivity of the pancreatic antibodies falls as the disease progresses in those with a diagnosis of T1DM. In this institute, to avoid diagnostic confusion with T1DM, individuals with the following characteristics are referred for genetic testing: a diagnosis of DM prior to 25 years of age, C-peptide > 0.6 ng/dl for five years after diagnosis of DM and negative islet autoantibodies (anti-glutamic acid decarboxylase [anti-GAD], anti-insulin antibodies and anti-tyrosine phosphatase antibodies).
The present study included individuals with subtypes of MODY in addition to HNF1A and GCK, which are the most prevalent. The study conducted by Shields et al. [4] included only European Caucasians of 1 to 35 years of age who had a confirmed genetic diagnosis of HNF1A, HNF4A or GCK-MODY. In the present study, the population recruited was in the same age range; however, patients with a confirmed molecular diagnosis not only of HNF4A, GCK and HNF1A but also PDX1, HNF1B, NEUROD1, INS and ABCC8 variants were included [5, 6]. All the variants included were pathogenic or likely pathogenic according to the ACMG/AMP criteria, with our group having experience in distinguishing between each one of these.
The cut-off point of the MODY calculator that resulted in the best sensitivity and specificity in the present study was 60% (sensitivity 96.1% and specificity 98.1%), followed by 40% (sensitivity 97.8% and specificity 96.4%), with accuracy being over 95% for all cut-off values above 40%. These values exceed that used in the Australian study [12] (cut-off of 25%) and that suggested by Shields et al. [4] (cut-off of 25% when comparing MODY with T2DM and 10% when comparing MODY with T1DM); however, in the present mixed-race population, the accuracy of these cut-off points was even higher.
A previous Brazilian study [14] conducted in Rio de Janeiro evaluated the prevalence of GCK- and HNF1A-MODY in an ethnically diverse population with suggestive clinical characteristics and analyzed what the impact would be of using the probability calculator for this purpose. Thirty-four patients aged ≤ 35 years at diagnosis of diabetes, with BMI < 30, negative anti-GAD and anti-IA2 antibodies, and a positive family history of diabetes in at least two generations, were analyzed. Patients with T1DM, a history of diabetic ketoacidosis, clinical signs of insulin resistance and DM of secondary causes were excluded from the analysis. In that study, cut-off points for the probability of MODY > 75% and > 62% were found for patients with HNF1A-MODY and GCK-MODY, respectively. Those findings suggest that higher cut-off points should be considered as indicative of a need to screen for MODY in non-Caucasian and mixed-race populations such as that of Brazil. Nonetheless, that study was limited by its small sample size, the absence of a control group, the fact that C-peptide was not measured and that genetic testing was only performed for two MODY genes; moreover, using only Sanger sequencing. Although the cut-off points were relatively higher (> 75% for HNF1A and 62% for GCK) compared to those found in the present study, comparison was not made with individuals with T1DM or T2DM as control groups; therefore, at that moment, the effectiveness of the MODY calculator could not be validated or evaluated in this population.
In view of these findings and considering the cost of molecular genetic testing, we would suggest using a cut-off point of 60% for the Brazilian population. Bearing in mind the socioeconomic differences in the different regions of Brazil, in those regions where testing is more economically viable and more accessible to the population, a cut-off limit of 40% could be used.
The limitations of the present study include the small number of patients with a diagnosis of T2DM and not having measured C-peptide or islet autoantibodies to define T1DM. In addition, the patients with T1DM and T2DM did not undergo genetic testing to exclude the possibility of MODY; therefore, we cannot affirm that there are no MODY patients classified as T1DM or T2DM in this study. A study involving a larger group of patients and with genetic testing of all the participants could be useful for identifying the most effective cut-off point for these groups.