Diabetes mellitus (DM) is a metabolic disease characterized by hyperglycemia resulting from insulin defects or resistance, which is associated with long-term organ dysfunction [1]. Recent studies [2-4] have demonstrated a positive correlation between glycemic control and lung function impairment in patients with DM. A meta-analysis including 3,182 diabetes patients and 27,080 control subjects found that patients with diabetes had decreased forced expiratory volume in 1 second (FEV1) and forced vital capacity (FVC) compared with a healthy control group [2]. One large-scale 5-year prospective study reported that individuals with diabetes had a 1.6-fold higher risk of restrictive lung function impairment than those without diabetes after adjusting for baseline lung function [4].

The relationship between glycemic control and lung function changes has been described in patients with DM in previous literature. Davis et al. [5] showed that declining lung function measures (FEV1, FVC) were consistently predicted by higher hemoglobin A1c (HbA1c) levels on follow-up in the Fremantile Diabetes study. The decline in lung function under poor glycemic control had been explained not only in DM but also in the prediabetes stages. Prediabetes, defined by HbA1c levels between 5.7% and 6.4% [1], had a negative correlation with both FVC (γ=-0.113, P<0.001) and FEV1 (γ=-0.079, P=0.003). A multivariable logistic regression model showed that restrictive patterns and FEV1 of <80% showed a positive correlation with HbA1c levels [6]. Moreover, epidemiological studies suggest that diabetes and hyperglycemia are associated with the development of several pulmonary diseases, including airway diseases and interstitial lung disease [7,8]. These studies suggest that the metabolic changes induced by hyperglycemia affect the lung, consistent with the pathologic autopsy findings showing involvement of the alveolar capillaries and arterioles, with hyalinosis, plasmorrhagia, and insudation, in diabetic microangiopathy [9]. The common links between diabetes and pulmonary diseases are systemic inflammatory pathways, including those involving proinflammatory cytokines such as tumor necrosis factor-α and interleukin-6, which induce insulin resistance [10,11]. Moreover, obesity, an important risk factor for DM, is considered to activate an inflammatory cascade that could affect lung function [7,12].

However, evidence regarding the association between lung function and glycemic control status in healthy individuals without prediabetes or DM is lacking. The prevalence of DM increases with age and is 34% in women and 14% in men among patients aged 20 to 39 years, whereas it is approximately 50% in those aged >80 years [13]. However, long-term follow-up studies are limited for the development of new-onset diabetes mellitus (NODM) in middle-aged adults. As the development of NODM is increasing in these patients, it is becoming increasingly important to determine the future risks of developing DM. Therefore, this study was designed to analyze HbA1c after a 6-year follow-up according to various lung function test results in healthy individuals, exploring the influence of lung function changes on glycemic control.

This study described an interesting negative association between the FEV1/FVC ratio and future DM risk in healthy individuals. Subjects with an FEV1/FVC ratio of 78 to 82 had a 40% lower risk of DM diagnosis in a 6-year follow-up compared with the reference group (FEV1/FVC ratio ≥86) after adjusting for age, BMI, and sex. However, percent predicted FVC, FEV1, FEV1/FVC ratio, and FEF_25%−75% showed a negative linear correlation with baseline HbA1c. During the 6-year interval, a negative correlation was still observed in the plot of changes in FEV1/FVC ratio or FEF_25%−75% versus HbA1c. That is, certain ranges of the FEV1/FVC ratio (78 to 82) could predict a decreased future risk of DM, and HbA1c increased with decreasing lung function parameters at baseline in non-diabetes individuals. Moreover, baseline HbA1c increased with changes in the FEV1/FVC ratio or FEF_25%−75% decrease for 6 years.

Several previous longitudinal studies evaluated the risk of future type 2 DM according to different baseline pulmonary function test results in non-diabetes subjects [16-20]. The consistent common findings of these studies were the increased risk of DM in the lowest quartile of predicted FEV1 (%) or FVC (%) compared with that in the highest quartile. They observed the incidence of DM during 9 to 37 years of follow-up using various diagnostic methods of DM. The diagnostic method was subjects’ report [18], fasting blood glucose level [19,20], or international classification of disease codes [16]. Instead of using these parameters to diagnose DM, Lazarus et al. [17] measured insulin levels and the fasting insulin resistance index to define insulin resistance. They found that baseline FVC, FEV1, and mid-expiratory flow rate were negatively associated with future insulin resistance after adjusting for baseline age, BMI, fat distribution pattern, and cigarette smoking [17]. Moreover, insulin secretion was also shown to be correlated to lung function parameters in previous studies. Alicandro et al. [21] reported that lower fasting blood C-peptide levels were associated with lower FEV1 values. Nezer et al. [22] also demonstrated that there was a negative correlation between baseline FEV1 and glucose level after a 2-hour oral glucose tolerance test (OGTT) and higher insulin secretion during pulmonary exacerbations in patients with cystic fibrosis.

Contrary to this, the results of our study revealed a decreased risk of NODM associated with an FEV1/FVC ratio of 78 to 82 compared with higher reference ranges (FEV1/FVC ratio ≥86) which could be explained by our study’s distinct population. We excluded diabetes patients according to HbA1c at baseline, who were mainly middle-aged with some older adults aged >60 years. These inclusion criteria would have strictly excluded hyperglycemic individuals compared with other studies, which defined DM based on self-reporting. Moreover, subjects who were taking glucose-lowering agents might have been missed at the 6-year follow-up because of the definition of DM based on HbA1c. In fact, the proportion of participants diagnosed with DM after the 6-year follow-up was 0.9% in our study, which is smaller than that of 2.2% after a 5-year follow-up among individuals from another health check-up program in the Republic of Korea as described in another report [23]. However, the main strength of this study is its large population cohort recruited from one tertiary referral university hospital. We included 17,568 subjects who had serial lung function test results with 6-year intervals performed by trained technicians according to international guidelines. To the best of our knowledge, this is the largest study population that has been used to predict future DM risk according to different ranges of lung function in non-diabetes individuals.

The occurrence of NODM in the young population has been recently gained much focus [24]. Because only few studies stratified the risk of NODM by age groups, our study could suggest new real-world evidence based on data from routine health check-ups. Data obtained from regular health check-up programs have several advantages compared with other big data regarding a variable number of clinical parameters and an easily tracked and consistent lung function testing protocol throughout the study. Moreover, we used NODM, a clear primary outcome for our analysis. These factors had enabled us to successfully identify the associations between variable clinical characteristics and NODM diagnoses.

One cross-sectional study explained the impaired glucose metabolism was an independent mechanism of airway resistance in patients with obesity and asthma. Karampatakis et al. [25] had reported that asthmatics with insulin resistance had significantly lower values of bronchial hyperresponsiveness (BHR) than those without insulin resistance. They described that obesity alone was not a causal factor in BHR because BMI was not associated with changes in the value of the provocation dose of mannitol required to achieve 15% drop in FEV1 [25,26]. Kim et al. [27] also found that among the 1,058 healthy subjects, 33 (3.1%) had BHR based on methacholine bronchial provocation test, and insulin resistance and BMI were the significant risk factor for BHR after adjusting for age in women. At the molecular level, the effect of hyperinsulinemia on airway smooth muscle contractility by mast cells [28,29], to alter small airway parasympathetic innervation [30], or to increase transforming growth factor-β activity in the airway [31] were suggested as possible mechanisms for the association between BHR and insulin resistance. Although we could not suggest an exact mechanism for the link between airway limitation and future DM risk in healthy subjects without diabetes, the impairment of glucose metabolism would be one of the important factors contributing to airflow resistance, based on these studies and our results.

However, evidence regarding the association between NODM and airflow obstruction from long-term follow-up studies is lacking. An FEV1/FVC ratio of <70 to 75 is defined as airway obstruction for the diagnosis and treatment of chronic obstructive pulmonary disease (COPD) according to international guidelines [32]. FEF_25%−75% is a forced mid-expiratory flow, known to be the most sensitive measure of airflow in the peripheral airways, associated with small airway diseases [33]. Impaired FEF_25%−75% in asymptomatic subjects has been defined as a surrogate marker for the early onset of asthma, BHR, or lower airway inflammation [34]. Moreover, the FEF_25%−75% value was suggested as a predictor for COPD development in patients with normal lung function in a recent analysis [35]. Because the participants included in this study were healthy middle-aged individuals from a health promotion center, we included FEF_25%−75% in analyzing the association between baseline lung function and future risk of DM to detect early small airway dysfunction. In previous studies, the baseline FEV1/FVC ratio did not show a statistically significant correlation with the diagnosis of DM [16,20]. The negative linear correlation between the FEV1/FVC ratio and future DM risk is a novel finding. Moreover, this is also the first trial to perform a regression analysis using FEF_25%−75% regarding the risk of NODM diagnosis.

The crude risk of DM diagnosis was 1.6 (95% CI, 1.1 to 2.5; P=0.03) in the lowest quartile (Q1) of the FEV1/FVC ratio compared with the reference quartile (Q4) (Table 3). However, this correlation was not observed after adjusting for compounding variables, including age and sex (model 1) or age, sex, and BMI (model 2). The risks were decreased in models 1 and 2 and were lowest in Q2 compared with reference Q4. Additional adjustment for history of pulmonary Tbc and asthma, which could affect airflow obstruction, showed the same result (data not shown). Therefore, the second quartile FEV1/FVC ratio would be an independent negative factor for future DM diagnosis after adjusting for older age, male sex, higher BMI, history of pulmonary Tbc, and bronchial asthma. This result is quite interesting because subjects in Q2 had higher proportion of hypertension, higher HbA1c, and slightly higher triglyceride and LDL-cholesterol levels than those in Q4 at baseline. Moreover, the negative linear correlation between baseline lung function parameter and HbA1c could describe uncovered equations in non-diabetes healthy individuals.

There are several limitations to our study. The first limitation is the relatively short follow-up duration and absence of annual HbA1c values. Zaigham et al. [20] analyzed the hazard ratios of DM according to lung function over a 30-year follow-up period and showed that there was increased future DM risk in the latter years compared with that in the earlier phase. Moreover, the prevalence of DM increases significantly with age in the general population [36], and a 5-year follow-up period would have underdiagnosed those at risk for DM. While HbA1c levels can reflect the blood glucose status of individuals in the preceding 2 to 3 months [37], an analysis of annual follow-up HbA1c levels could reflect hyperglycemic changes more accurately. Instead of the cross-sectional study design employed in this study, a prospective longitudinal observational study design analyzing the risk of DM according to the FEV1/FVC ratio would be needed in future research. Another possible limitation of this study was that we missed adjusting for cigarette smoking history in the multivariate regression analysis. It is possible that the slightly decreased baseline FEV1 and FEV1/FVC ratio values in the male participants could have been affected by cigarette smoking. Although we included male sex as a confounding factor in the risk of DM, smoking history should be considered in future analyses. The third limitation is the restricted age group at baseline since 91.4% of the 17,568 individuals enrolled in our study were under the age of 60. However, in terms of predicting lung function deterioration in the future, this may be a clinically appropriate age range. Furthermore, this study was limited to patients who underwent a health check-up. There was also a limitation in the data that we were able to collect from the health-promotion center. Since the subjects visited the center at intervals of approximately 3 to 24 months, we could not define the exact timing of NODM diagnosis and HbA1c differences during the 6-year period. In addition, that we did not consider physical activity as a risk factor for NODM risk analysis could be a limitation. Epidemiological studies have demonstrated that low levels of physical activity and fitness are prominent, independent, and modifiable risk factors for the development of type 2 DM [38-40]. Finally, the single-center cohort design of this study was a limitation. However, since we performed a comparison of the parameters after a 6-year period, we considered the reliability of the laboratory data to be high enough that this may be considered a strength.

In conclusion, our study described that the baseline FEV1/FVC ratio of 78 to 82 was associated with a 40% lower risk of future DM diagnosis than a higher FEV1/FVC ratio of ≥86 after adjusting for age, sex, and BMI. In non-diabetes individuals, decreased FEV1 (%), FVC (%), FEV1/FVC ratio, and FEF_25%−75% were associated with an increased HbA1c level in linear regression analysis. These findings could enable the determination of the risk of new-onset diabetes after 6 years based on lung function parameters in healthy subjects, which could identify the at-risk population for DM through further large-scale research.