2024 Volume 71 Issue 11 Pages 1093-1096
Elevated Fulminant Index (FI), [plasma glucose (PG)/glycosylated hemoglobin A1c (HbA1c)], was reportedly a sensitive index to differentiate fulminant type 1 diabetes (FT1D) from non-fulminant T1D (nFT1D). Aim of this study was to describe a better, but simpler index of FT1D. 49 and 52 patients with FT1D and nFT1D, respectively, were registered, and the discriminating ability of the rounded, normalized ratio, [PG (mmol/L) – 5.0]/[HbA1c (%) – 5.0], and the original ratio, [PG (mmol/L)]/[HbA1c (%)], was compared. Normalizing the ratio significantly raised its accuracy: area under the curve for receiver operating curve, AUROC (95%CI), 0.927 (0.858–0.964) and 0.851 (0.763–0.910), respectively, with and without the normalization (p < 0.01). Rounding of the figure into [PG (mmol/L) – 5.0]/[HbA1c (%) – 5.0] did not significantly sacrifice the discriminating ability of the index. Namely, the optimal cut point of rounded and normalized GAR, 10.0, showed 89.8% sensitivity. In conclusion, rounded, normalized (rn) GAR ≥10 (the rounded optimal cut-off) could be used for the snap diagnosis of FT1D.
Fulminant type 1 diabetes (FT1D) is a subtype of type 1 diabetes first reported by Imagawa et al. in 2000 [1]. Since then, many case reports on this entity have been accumulated worldwide, and early recognition of the condition and swift placement of the patients on the high-alert, critical care unit has been established as the key for the successful treatment. This is because development of diabetic ketoacidosis (DKA) is so rapid and its degree so severe in FT1D [2, 3]. Regarding incidence of FT1D, it has been extremely rare in Caucasians [2, 4-6] so that it is suggested that different genetic/environmental interactions might operate in the etiology of T1D between Caucasians and Asians [5-7]. We are afraid that apparent rarity of FT1D in Caucasians [2, 5, 6] might be in part due to simple “non-diagnosis” of the condition. Notably, there is no numerical, clinically usable index for differentiation of FT1D from nonfulminant type 1 diabetes (nFT1D). An exception was work of Qiu et al. [8] in which they reported that “glucose/HbA1c” of the initial sample could be employed as a tool for rapid, efficient and easy differentiation of FT1D from nFT1D. They named it as the Fulminant Index. In this study, we substantially improved FI for accuracy and reliability in the diagnosis of FTID.
This is a retrospective, multi-center-collaborative study on Japanese adult (>20 years) cases with DKA: the patients complicated with pregnancy were excluded. Diagnosis of FT1D was made according to the 2012 recommendation of Japan Diabetes Society [9]. First, patients with FT1D were collected by PubMed search. Namely, two key words, “fulminant type 1 diabetes” and “case” were entered to PubMed on 9/1/2022 and the literature published between 2012 and 2022 were checked. 114 case reports were hit and each case was checked for compatibility of the diagnostic criteria of F1TD described above [9]. 23 of the 114 publications were excluded because of non-compatibility of the diagnosis and/or lacking the pretreatment laboratory data such as HbA1c, serum K+ and blood pH. 50 reports (55 cases) on non-Japanese subjects were further excluded. The rest (26 cases) were used in this study. Additionally, the data from 15 cases with F1TD were collected by the literature search of papers published in Japanese at Igaku-Chuo-Zasshi (https://www.jamas.or.jp/shusaishi/view/?jid=J03170). This was done on 4/16/2023 by entering the same key words as above for PubMed and we checked that each case hit by Igaku-Chuo-Zasshi was different from those hit by PubMed search. 4 FT1D cases diagnosed and treated at Aizawa Hospital and Nagano Chuo Hospital, all of which fulfilled the above-described diagnostic criteria were also included. For comparison, the data from 52 patients with nFDKA were collected at Aizawa Hospital, Nagano Chuo Hospital, Asama General Hospital and Shinshu University Hospital, between 2020 and 2022.
Characteristics of the study subjectsAge, sex and the essential laboratory data of patients with FT1D and nFDKA are shown in Table 1.
| Variables | FT1D | nFDKA | p value* |
|---|---|---|---|
| N | 49 | 52 | N.A. |
| Age, years | 48 (36–63) | 55 (38–70) | 0.35 |
| Sex, M/F | 23/26 | 32/20 | 0.14 |
| HbA1c, % HbA1c, mmol/mol |
6.78 (6.15–7.30) 50 (43–56) |
10.40 (8.78–12.88) 90 (72–117) |
N.A. |
| PG, mmol/L | 46.3 (31.6–74.1) | 39.0 (22.9–51.5) | <0.01 |
| K+, mmol/L | 6.0 (5.0–7.0) | 5.0 (4.6–5.9) | 0.01 |
| Indices | |||
| nGAR | 30.61 (14.14–53.28) | 5.78 (3.28–9.12) | <0.01 |
| PG/HbA1c | 7.1 (4.6–10.9) | 3.6 (1.9–4.8) | <0.01 |
| K+/HbA1c | 0.900 (0.766–1.005) | 0.500 (0.352–0.642) | <0.01 |
Data are median (interquartile range). FT1D, fulminant type 1 diabetes; nFDKA, nonfulminant diabetic ketoacidosis; *, F1TD vs. nFDKA by Wilcoxon rank-sum test or x2 test. nGAR, normalized glucose/A1c ratio. Unit for PG/HbA1c, mmol/L/%. N.A., not applicable. HbA1c level <8.7 % is one of the required conditions for making diagnosis of F1TD [9], and therefore, statistical difference between the two groups is a priori.
As the novel index of fulminancy, we coined “normalized glucose/A1c ratio (nGAR)”:
nGAR = [plasma glucose (PG) (mmol/L) – 5.5*]/[glycosylated hemoglobin A1c (HbA1c) (%) – 5.1†], where 5.5* and 5.1† are the median FPG and HbA1c, respectively, obtained from 1,125 apparently healthy Japanese adults [10].
Statistical analysisUnivariate logistic regression analysis was performed employing F1TD as the objective variable with nFDKA as the control. In Model 1, nGAR was entered as an explanatory variable. In Model 2 and 3, plasma glucose (PG)/HbA1c and K+/HbA1c were employed, as explanatory variables. K+/HbA1c and PG/HbA1c were taken into the analysis because the two were the best and second to the best discrimination of FT1D and nFDKA in Qiu’s study [8]. ROC curves were drawn for each explanatory variable and the optimal cutoff point was determined as the point where Youden index was maximum. Also, area under the curve (AUC) was calculated for each index and compared by DeLong’s test. For clinical use, rounding of the index was performed in search of the easily memorable cut off. P < 0.05 was considered statistically significant. JMP Pro 16.0 was used for statistical analysis.
The study at the 5 hospitals was inclusively approved by the ethics committee of Aizawa Hospital (2019-027).
Characteristics of the study subjectsThe age and sex distribution were not significantly different between FT1D and nFDKA patients. On the other hand, PG and serum K+ were significantly higher in the former than in the latter. nGAR, PG/HbA1c and K+/HbA1c were also significantly higher in the former.
Comparison of the discriminatory efficiency of the indices for FT1D and nFDKAAUC of ROC (95%CI) was 0.927 (0.858–0.964) for nGAR, 0.851 (0.763–0.910) for PG/HbA1c. The former was significantly greater than the latter (p = 0.01). The optimal cutoff point for nGAR was 10.04 which yielded 89.8% sensitivity and 82.7% specificity. Corresponding values for PG/HbA1c was 4.86, and 85.0% and 82.7%, respectively, and the values were lower for PG/HbA1c except that the specificity was identical for the two indices.
Clinical applicationTo maximize practicability of the index, we tried the following rounding.
Rounded nGAR = [PG (mmol/L) – 5.0]/[HbA1c (%) – 5.0], and also rounded the maximal Youden index, 10.04, to 10.00. As shown in Table 2, this modification had minimum sacrificing effect on the discriminatory ability of the index.
| Model | AUC for ROC (95%CI) | p value | Optimal cutoff point | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|---|---|
| 1. nGAR | 0.927 (0.858–0.964) | 1 vs. 2, 0.01 | 10.04 | 89.8% | 82.7% | 86.1% |
| 2. PG/HbA1c | 0.851 (0.763–0.910) | 2 vs. 3, <0.01 | 4.86 | 85.0% | 82.7% | 78.2% |
| 3. K+/HbA1c | 0.892 (0.809–0.941) | 3 vs. 1, 0.11 | 0.680 | 86.5% | 81.6% | 84.2% |
| 4. Rounded nGAR | 0.926 (0.856–0.963) | 4 vs. 1, 0.21 4 vs. 2, <0.01 |
10.0 | 89.8% | 81.7% | 79.4% |
nGAR, normalized glucose/A1c ratio; AUC, area under the curve; p values for statistical difference of AUC, Rounded nGAR was defined as, Rounded nGAR* = [PG (mmol/L) – 5.0]/[HbA1c (%) – 5.0].
*If the unit of plasma glucose is mg/dL, rnGAR = {[PG (mg/dL) – 90]/18}/{HbA1c(%) – 5.0}
We significantly improved the FI reported by Qiu et al. [8] by introducing negative constants for the original equation. Resultant novel index, nGAR, was rounded for the sake of easy memorability. Plasma glucose and HbA1c should have been stayed at normal basal value, not zero, before onset of diabetes. Therefore, calculating glucose/HbA1c after subtracting such basal values, 5.5 mmol/L for glucose and 5.1% for HbA1c, gave us the ratio accurately reflecting the discordance between PG and HbA1c. Subtraction of basal, normal value to make the index with increased accuracy is the same as homeostasis model assessment beta cell function [11]. Of course, the basic assumption here was relatively slow rise of HbA1c compared to PG upon acute elevation of PG [12].
Practically, calculation of nGAR is so simple that it would be useful for any health professionals. To further improve clinical utility, we rounded constants in the equation and the optimal cut point. On the basis of the result shown in Table 2, for the bed-side snap diagnosis of FT1D, we propose rnGAR ≥10.
Nevertheless, any clinical guidelines cannot be free from exceptions. In the case of glucose/A1c ratio, the patients with severe anemia or preexisting dysglycemia would show unduly low or high HbA1c. HbA1c, at the time of diagnosis of FT1D, was reportedly higher than that of the conventional FT1D, in those on antibodies against immune checkpoint inhibitors [13]. Finally, exceptionally low PG (<16.5 mmol/L, <300 mg/dL) was found in 4 out of 161 patients with F1TD upon the nationwide survey in Japan [14]. Under these conditions, nGAR might not be a reliable index of fluminancy of metabolic derangement any longer. One should keep these facts in mind to avoid overlooking FT1D.
In conclusion, the original FI of Qiu et al. [8] was modified to improve its discriminatory efficiency. Simple subtraction of the expected median from PG and HbA1c caused significant improvement of the index. rnGAR thus obtained, with rounding, is useful as a quantitative index for FT1D, for healthcare professionals at large (Graphical Abstract). Wide-spread utilization of it would save lives of previously not diagnosed FT1D.
Dr. M Komatsu is a member of Endocrine Journal’s Editorial Board.
FundingThis work was not supported by any organization.
Duality of InterestNo potential conflicts of interest relevant to this article were reported.
Author ContributionsJ.M., T.M, and T.A. made the study design. H.Y. conducted analyses and wrote the first draft of the manuscript. T.K., E.N., M.N., A.S., M.K., and K.Y. contributed to discussion of the results and critical revision of the manuscript. All authors approved the final version of the manuscript. T.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
