Calculated estimation of the metabolic clearance rate of insulin measured by glucose clamp examination in out-patient clinical practice

Abstract

Purpose

A subpopulation of Japanese patients with type 2 diabetes mellitus (T2DM) who have elevated insulin clearance (IC) exists. We tested our hypothesis that it is possible to estimate IC using common and simple test results collected in routine clinical practice.

Methods

We recruited patients with newly diagnosed, treatment-naïve T2DM and measured the metabolic clearance rate of insulin (MCRI) determined by a hyperinsulinemic–euglycemic clamp examination. Multivariable regression analysis was performed with body mass index (BMI), serum uric acid (UA), and fasting plasma insulin (F-IRI) which were independently associated with IC increase in our previous reports as explanatory variables to calculate a prediction equation for MCRI.

Results

We enrolled 101 patients in this study. Because MCRI is not normally distributed, we calculated the logarithmically transformed estimated Log₁₀MCRI as a prediction formula for IC. Multivariable regression analysis showed that Log₁₀BMI (β =  − 0.3257, P < 0.001), UA (β =  − 0.1834, P = 0.0081), and Log₁₀F-IRI (β =  − 0.4367, P < 0.001) were significant independent factors for Log₁₀MCRI. The regression equation was as follows: estimated Log₁₀MCRI =  − 0.5421 × Log₁₀BMI − 0.0167 × UA − 0.1792 × Log₁₀F-IRI + 3.8251 (r = 0.7677, R² = 0.5894, P < 0.001).

Conclusion

IC can easily be predicted using BMI, UA, and F-IRI which are common and simple test results collected in routine clinical practice.

Introduction

In our previous study, we reported that there is a subpopulation of Japanese patients with type 2 diabetes mellitus (T2DM) who have elevated insulin clearance (IC) and that their body mass index (BMI), serum uric acid (UA), and fasting plasma insulin (F-IRI) levels were significantly and independently associated with elevated IC [1]. It is important to understand an individual’s IC rate in T2DM treatment. However, the hyperinsulinemic–euglycemic clamp test must be performed to accurately evaluate IC, and this test is difficult to perform in daily clinical practice. In this study, we tested our hypothesis that it is possible to estimate IC using common and simple test results collected in routine clinical practice.

Methods

Study patients and protocol

Because this study is a supplementary analysis to our previously reported study [1], the study protocol can be referred to in the published report. The metabolic clearance rate of insulin (MCRI), an indicator of IC, was calculated according to the following formula: MCRI (ml/min/m²) = (insulin infusion rate at steady state)/(steady-state serum insulin). This study was conducted in accordance with the Declaration of Helsinki, and the study protocol was approved by the Human Ethics Review Committee of Jinnouchi Hospital, Kumamoto, Japan (2018–3-4). This study was registered in the UMIN protocol registration system (ID: UMIN000032014).

Statistical analysis

The Shapiro–Wilk test was used to assess the normal distribution of continuous data. For parametric tests, the data that were not normally distributed were logarithmically transformed to achieve a normal distribution. The Pearson’s product–moment correlation coefficient was determined to investigate the correlation between Log₁₀MCRI and Log₁₀BMI, UA, and Log₁₀F-IRI. Multivariable regression analysis was performed with Log₁₀BMI, UA, and Log₁₀F-IRI as explanatory variables to calculate a prediction equation for Log₁₀MCRI. The estimated Log₁₀MCRI calculated from the prediction equation was exponentially converted back to a true number to derive the estimated MCRI (eMCRI). Receiver-operating characteristic (ROC) curve analysis was performed to determine the diagnostic ability of eMCRI to detect elevated IC. In our previous study [1], we demonstrated the usefulness of the diagnostic ability of F-IRI in detecting elevated IC. Therefore, in this study, we compared the diagnostic abilities of F-IRI and eMCRI for elevated IC. A P value < 0.05 denoted statistical significance. Statistical analyses were performed using SPSS version 23 (SPSS Inc., Tokyo, Japan).

Results

Subjects

We enrolled 101 patients in this study; 78.2% were men. Patients had a mean age of 54.1 years and a median BMI of 25.1 kg/m².

Shapiro–Wilk test

The Shapiro–Wilk test of BMI, UA, F-IRI, and MCRI showed that UA (P = 0.3715) was normally distributed, while BMI (P < 0.001), F-IRI (P < 0.001), and MCRI (P < 0.001) were not normally distributed. BMI, F-IRI, and MCRI, which were not normally distributed, were logarithmically transformed, and the Shapiro–Wilk test was performed again. Those results showed that Log₁₀BMI (P = 0.2987), Log₁₀F-IRI (P = 0.7936), and Log₁₀MCRI (P = 0.1902) were normally distributed.

Pearson’s product–moment correlation coefficient for Log₁₀MCRI

Log₁₀BMI (r = −0.66, P < 0.001), UA (r = −0.38, P < 0.001), and Log₁₀F-IRI (r = −0.70, P < 0.001) were significantly and negatively correlated with Log₁₀MCRI.

Multivariable regression analysis for Log₁₀MCRI

Table 1 shows the results of the multivariable regression analyses for MCRI. Log₁₀BMI (β =  − 0.3257, P < 0.001), UA (β =  − 0.1834, P = 0.0081), and Log₁₀F-IRI (β =  − 0.4367, P < 0.001) were significant independent factors for Log₁₀MCRI. The regression equation was as follows: estimated Log₁₀MCRI =  − 0.5421 × Log₁₀BMI − 0.0167 × UA − 0.1792 × Log₁₀F-IRI + 3.8251 (r = 0.7677, R² = 0.5894, P < 0.001).

Table 1.

Results of the multivariable regression analyses for Log₁₀ metabolic clearance rate of insulin

	Unstandardized coefficients		Standardized coefficients		Upper bound	P value
	β	SEM	B	Lower bound
Log10 body mass index	−0.5421	0.1448	−0.3257	−0.8294	−0.2548	< 0.001
Serum uric acid	−0.0167	0.0062	−0.1834	−0.0289	−0.0044	0.0081
Log10 fasting plasma insulin	−0.1792	0.0356	−0.4367	−0.2499	−0.1086	< 0.001

R² = 0.5894. The F test for a linear relationship was 46.4144 with P < 0.001. Variance inflation factors (used to detect multicollinearity between exploratory variables): Log₁₀ body mass index, 1.7868; serum uric acid, 1.0887; Log₁₀ fasting blood insulin, 1.7770. SEM, standard error of the mean

ROC curve analysis for elevated IC

ROC curve analysis was performed to determine the diagnostic ability of the eMCRI for elevated IC. eMCRI is an exponential conversion of the estimated Log₁₀MCRI. The results are shown in Fig. 1. The cut-off value of eMCRI for the presence of elevated IC was 744.4, and the area under the curve (AUC) was 0.935 (sensitivity 95%, specificity 77%, P < 0.001). In a previous study [1], we reported that the value of the F-IRI level for the presence of elevated IC was 5.5 μU/mL and that the AUC was 0.887 (sensitivity 88%, specificity 77%, P < 0.001). The diagnostic ability of eMCRI for the presence of elevated IC was significantly higher than that of the F-IRI level [difference in the AUC values: 0.0480, standard error (SE): 0.0212, 95% confidence interval (CI): 0.0066 to 0.0895, P = 0.0232].

Fig. 1.

Receiver-operating characteristic curve analysis of the F-IRI and eMCRI for detecting elevated IC. The diagnostic ability of eMCRI for the presence of elevated IC was significantly higher than that of the F-IRI level (difference in the AUC values: 0.0480, SE: 0.0212, 95% CI: 0.0066 to 0.0895, P = 0.0232). F-IRI fasting plasma insulin, eMCRI estimated metabolic clearance rate of insulin, IC insulin clearance

Discussion

This study demonstrated that it is possible to estimate IC using BMI, UA, and F-IRI. Additionally, the diagnostic ability of the eMCRI for the presence of elevated IC was very high. Because, we have reported that gender, age, and weight were not independently and significantly associated factors with elevated IC in our previous study, this estimation formula can be used independent of gender, age, and weight.

The estimation of MCRI is not completely indirect, because the eMCRI includes the F-IRI level, but the F-IRI was combined with BMI and UA to further improve the diagnostic ability for determining elevated IC. At this stage, it is difficult to prove why including uric acid levels as an explanatory variable improves the accuracy of eMCRI estimation, and this is considered a future issue.

In conclusion, IC can easily be predicted by the equation Log₁₀MCRI =  − 0.5421 × Log₁₀BMI − 0.0167 × UA − 0.1792 × Log₁₀F-IRI + 3.8251.

Funding

This research did not receive any specific grants from funding agencies in the public, commercial, or not-for-profit sectors.

Declarations

Conflict of interest

HJ has received honoraria from Novo Nordisk, Sanofi, AstraZeneca Pharmaceuticals, Astellas Pharma, Boehringer Ingelheim, Daiichi-Sankyo, Eli Lilly, Takeda, and Novartis Pharmaceuticals. SS has received honoraria from AstraZeneca Pharmaceuticals and Ono Pharmaceutical. The authors declare that no other potential conflicts of interest exist related to this study.

Ethical approval

This study was conducted in accordance with the Declaration of Helsinki, and the study protocol was approved by the Human Ethics Review Committee of Jinnouchi Hospital on March 26, 2018 (approval number: 2018–3-4). Informed consent or substitute for it was obtained from all patients for being included in the study.