Herein, we investigated the ability of TyG-BMI to predict all-cause mortality, developed a new index for predicting mortality using TyG-BMI, and evaluated its clinical usefulness in nonobese AAV patients. We obtained several interesting findings. Firstly, the AUC of TyG-BMI for all-cause mortality in the 78 nonobese AAV patients, tended to be statistically significant but those of all or obese AAV patients showed no similar trend of significance. Secondly, when the cut-off value of TyG-BMI for all-cause mortality was set as 187.74, all-cause mortality was more frequently identified in AAV patients with TyG-BMI ≥ 187.74 than those with TyG-BMI < 187.74. Thirdly, a new index using TyG-BMI for all-cause mortality (NITGB) was developed using variables that were with a P-value < 0.1 in the Cox hazards model analyses for predicting all-cause mortality. Fourthly, when the cut-off value of NITGB for predicting all-cause mortality was set as 27.36, the cumulative mortality rate in nonobese AAV patients with NITGB ≥ 27.36 was significantly higher than that in those with NITGB < 27.36. Therefore, it is concluded that TyG-BMI and NITGB at AAV diagnosis could be useful in predicting all-cause mortality, and furthermore, the predictive ability of NITGB is greater than that of TyG-BMI in nonobese AAV patients.
Only nonobese AAV patients were included herein. BMI is known to be closely associated with mortality in the general population. In particular, it shows a U shape, which is characterised by roughly opposite results between the nonobese and obese people.20 Since BMI is proportional to the probability of all-cause mortality in individuals with BMI ≥ 23.0 kg/m2, the mortality-reflected effect of BMI itself is added to the predictability of all-cause mortality. TyG, which may interfere with interpreting the predictability of all-cause mortality of TyG-BMI and leave its lower reliability. Whereas, since BMI is inversely correlated with the probability of all-cause mortality in individuals with BMI < 23.0 kg/m2, BMI may highlight the predictability of all-cause mortality of TyG-BMI in AAV patients. Actually, when all patients were analysed in this study, the AUC of TyG-BMI for all-cause mortality was not significant, whereas it showed a tendency to be statistically significant in 78 non-obese AAV patients in the ROC curve analysis. These results indicated that the risk of all-cause mortality associated with TyG-BMI is meaningful in non-obese people but it is not in overweight and obese people (Fig. 1). This finding is consistent with findings of other studies which showed that non-obese people have a higher risk of TyG-BMI related diabetes13 and non-alcoholic fatty liver disease15 than overweight and obese people. However, we were not able to provide evidence regarding whether the presence or absence of a U shape exist in the association between the risk of all-cause mortality and BMI in our data because the number of mortality case was small.
Since TyG-BMI consists of fasting plasma TG, fasting plasma glucose, TyG, and BMI, we assessed the abilities of each component for predicting all-cause mortality and compared them with TyG-BMI in nonobese AAV patients. First, in the ROC curve analysis, the AUCs of fasting plasma TG, fasting plasma glucose, TyG and BMI were 0.547 (P = 729), 0.521 (P = 0.878), 0.570 (P = 0.603), and 0.688 (P = 0.162), respectively. There was no trend of the significant association between each component composing an equation of TyG-BMI and all-cause mortality (Suppl Fig. S2). Next, in the univariable Cox hazards model analysis, fasting plasma TG (HR 1.001, P = 0.908), fasting plasma glucose (HR 1.010, P = 0.509), TyG (HR 1.831, P = 0.536), and BMI (HR 1.541, P = 0.186) were not associated with all-cause mortality. Whereas TyG-BMI ≥ 187.74 (HR 6.264, P = 0.045) was independently and significantly associated with all-cause mortality in nonobese AAV patients. Therefore, to cope with the possibility of all-cause mortality during follow-up in nonobese AAV patients, attention should be carefully paid to the predictability of TyG-BMI over the cut-off for predicting mortality at the time of AAV diagnosis.
In this study, in order to meet the clinical need for a new index for predicting all-cause mortality, the cut-off value of TyG-BMI with a P-value < 0.1 was obtained from the ROC curve. To overcome this statistical limitation, we also analysed the implication of TyG-BMI for presupposing all-cause mortality using two more cut-offs, the highest tertile, and quartile, that are frequently and clinically applied. In terms of the highest tertile of TyG-BMI, the lower limit of the highest tertile was set as 179.28. When nonobese AAV patients were partitioned into two groups, there was no significant difference in the cumulative mortality rates between patients with TyG-BMI ≥ 179.28 and those with TyG-BMI < 179.28 (P = 0.276) (Suppl Fig. S3A). In terms of the highest quartile of TyG-BMI, the lowest value of the highest quartile was calculated as 183.57. When nonobese AAV patients were divided into two groups, patients with TyG-BMI ≥ 183.57 tended to exhibit a significantly higher cumulative mortality rate than those with TyG-BMI < 183.57 but it did not reach statistical significance (Suppl Fig. S3B). Therefore, it is concluded that 187.74 obtained from the ROC curve is the suitable cut-off value for predicting all-cause mortality in nonobese AAV patients.
TyG is an index composed of fasting plasma TG and fasting plasma glucose and has been well known as an index that effectively reflects IR and overall metabolic status in individuals.25 However, Er et al. found that the new index of TyG-BMI, which is formed by combining the TyG index with BMI, can better reflect IR status than the TyG index.13,14 Our study results provide clinically supporting evidence for the biologically plausible hypothesis that IR plays an important role in the AAV patients. Although the underlying mechanism of the relationship between TyG-BMI and all-cause mortality in AAV patients is unclear, it may be related to IR. TyG as a surrogate marker of IR is associated with metabolic disorders including T2DM. Consequently, it ultimately increases the risk of all-cause mortality by enhancing the possibility of their systemic complications including CVA and CVD.26 However, when the association between TyG-BMI and CVA or ACS was investigated using the ROC curve and univariable Cox hazards model analysis, the AUCs of TyG-BMI for CVA and CVD were not statistically significant (Suppl Fig. S1), and furthermore, TyG-BMI was significantly associated with neither CVA (HR 1.026, P = 0.315) nor CVD (HR 1.033, P = 0.293).
It is inferred that the pathological association between IR increased by TyG-BMI and the clinically significant occurrence of CVA or CVD in nonobese AAV patients was not as robust as that in AAV patients regardless of BMI. In fact, when the univariable Cox analysis was performed in all 141 AAV patients, TyG-BMI showed a tendency to be associated with the occurrence of CVD but it was not statistically significant (HR 1.014, P = 0.110). Therefore, the mechanism by which TyG-BMI could predict all-cause mortality in nonobese AAV patients could not be explained solely by increased IR and subsequently the augmented possibility of CVA, and CVD.
In Table 2, in the multivariable Cox analysis, in addition to TyG-BMI ≥ 187.74, both age and BVAS were also associated with all-cause mortality in nonobese AAV patients. First of all, variables of age and the male sex are ready-established conventional risk factors for all-cause mortality.8 In the previous studies, we demonstrated that elderly AAV patients, as well as male AAV patients, had a significantly higher rate of all-cause mortality.10,11 However, a variable of age could predict all-cause mortality independently but that of the male sex could not in this study. It is assumed that these results might be owing to a relatively small proportion of male patients compared to females. Furthermore, the inclusion of only AAV patients with BMI < 23.0 kg/m2 might have influenced the results by diminishing the effect of obesity on all-cause mortality. On the other hand, BVAS is the most widely used index for assessing the activity of AAV.6 In addition to a role to reflect the cross-sectional activity of AAV, BVAS has been considered associated with all-cause mortality, in particular, cardiovascular disease-related mortality in AAV patients.27,28
Given these concepts, in this study, we assigned weights to age, BVAS, and TyG-BMI ≥ 187.74 by referring to the slope of the multivariable Cox analysis, and developed a new index using TyG-BMI for predicting all-cause mortality in nonobese AAV patients. We found two advantages of NITGM compared to TyG-BMI. One is that NITGM exhibited a more robust ability for predicting all-cause mortality than TyG-BMI itself because the abilities of age and BVAS for predicting all-cause mortality were added to that of TyG-BMI. The other is that NITGM includes three risk factors for all-cause mortality in a variety of areas: metabolic disorders-related cardiovascular (TyG-BMI), conventional (age), and AAV-specific (BVAS) risk factors. Finally, we found that the independent ability of NITGM ≥ 27.36 to predict all-cause mortality tended to be stronger than that of TyG-BMI ≥ 187.74 by the comparative analysis of the cumulative rates of all-cause mortality between the two groups (Fig. 4).
For the first time, we demonstrated the abilities of TyG-BMI and a new index using TyG-BMI (NITGB) to predict all-cause mortality in nonobese AAV patients. Our findings provide evidence that these indices are reliable markers for the early identification of individuals at high risk of mortality. Given the relatively high mortality rate of AAV, it is believed that this study has an advantage in that it enabled the development of biomarkers or indices at the time of AAV diagnosis for predicting all-cause mortality during follow-up in nonobese AAV patients. Additionally, the evaluated indices can be flexibly applied to AAV patients with ethnic and geographical differences by applying a method for deriving the cut-offs of TyG-BMI and NITGB rather than suggesting fixed values.
This study had several limitations. Firstly, the number of patients was too small to derive statistically sufficient significance. Thus, we adjusted the significance level (P-value < 0.05 to < 0.1) in the ROC curve and the Cox analyses. Secondly, because of the retrospective study design, we could not fully control for subclinical confounding factors that could affect not only TyG-BMI and NITGB at diagnosis but also all-cause mortality during follow-up. Moreover, we could not provide the serial data regarding TyG-BMI and NITGB from the time of AAV diagnosis to either the date of death or the last visit. We believe that a future prospective study with a larger number of nonobese AAV patients will validate our results and suggest the possibility of applying them to AAV patients in real clinical practice by showing the dynamic information on the association between either TyG-BMI or NITGB and all-cause mortality.