Respiratory Outcomes of Insulin Use in Patients with COPD: A Nationwide Population-Based Cohort Study

Acute exacerbations of chronic obstructive pulmonary disease (COPD) with severe hyperglycemia may require insulin to lower glucose levels in people with coexisting type 2 diabetes (T2D) and COPD. We conducted this study to examine the risk of hospitalization for COPD, pneumonia, ventilator use, lung cancer, hypoglycemia, and mortality with and without insulin use in people with T2D and COPD. We adopted propensity-score-matching to identify 2370 paired insulin users and non-users from Taiwan’s National Health Insurance Research Database between 1 January 2000 and 31 December 2018. Cox proportional hazards models and the Kaplan–Meier method were utilized to compare the risk of outcomes between study and control groups. The mean follow-up for insulin users and non-users was 6.65 and 6.37 years. Compared with no insulin use, insulin use was associated with a significantly increased risk of hospitalization for COPD (aHR 1.7), bacterial pneumonia (aHR 2.42), non-invasive positive pressure ventilation (aHR 5.05), invasive mechanical ventilation (aHR 2.72), and severe hypoglycemia (aHR 4.71), but with no significant difference in the risk of death. This nationwide cohort study showed that patients with T2D and COPD requiring insulin therapy may have an increased risk of acute COPD exacerbations, pneumonia, ventilator use, and severe hypoglycemia without a significant increase in the risk of death.


Introduction
Genetic factors, early-life events, infections, smoking, and environmental pollutants are the risk factors for developing chronic obstructive pulmonary disease (COPD) [1][2][3]. These factors can lead to lung damage or remodeling of the airways, pulmonary parenchyma, and vasculature, resulting in progressive airway limitation, dyspnea, exercise intolerance, restricted daily activity, and premature mortality [1,2]. Unstable COPD or infections can lead to episodic acute exacerbations. Each exacerbation contributes to pulmonary function deterioration and patient suffering [1,4,5]. The number of people with COPD worldwide excluding unsuitable patients, there were 3490 insulin users and 5667 nonusers. Finally, we identified 2370 insulin users and 2370 nonusers through 1:1 matching of the propensity scores ( Figure 1). The basic characteristics, comorbidities, and medication were well matched between the study group and the control group with SMD values <0.1 (Table 1). In the comparison cohorts of the insulin users and nonusers, the male proportion was 62.45% and 62.28%, respectively; the mean (SD) age was 61.02 (11.11) and 60.73 (11.21) years, respectively; the mean follow-up time was 6.65 (4.34) and 6.37 (4.29) years in each case. The basic characteristics, comorbidities, and medication were well matched between the study group and the control group with SMD values <0.1 (Table 1). In the comparison cohorts of the insulin users and nonusers, the male proportion was 62.45% and 62.28%, respectively; the mean (SD) age was 61.02 (11.11) and 60.73 (11.21) years, respectively; the mean follow-up time was 6.65 (4.34) and 6.37 (4.29) years in each case. The cumulative incidence of hospitalization for COPD (Log-rank test p-value < 0.001), bacterial pneumonia (Log-rank test p-value < 0.001), and IMV (Log-rank test p-value < 0.001) were significantly higher in insulin users than in non-users ( Figure 2).

Dose-Response Analysis
The cumulative duration (<90, 90-179, >179 days) of insulin use had significantly increased risks of hospitalization for COPD, bacterial pneumonia, and IMV than insulin

Dose-Response Analysis
The cumulative duration (<90, 90-179, >179 days) of insulin use had significantly increased risks of hospitalization for COPD, bacterial pneumonia, and IMV than insulin no-use, and longer cumulative insulin use appeared to increase the risk of these outcomes (Table 3).

Discussion
This nationwide cohort study demonstrated that patients with coexisting T2D and COPD requiring insulin therapy might increase the risk of hospitalization for COPD, bacterial pneumonia, ventilation use, and severe hypoglycemia; however, it did not increase the risk of death. Longer cumulative insulin use had association with increased risk of hospitalization for COPD, bacterial pneumonia, and invasive mechanical ventilation than insulin no-use.
Acute attacks of COPD often require hospitalization and can lead to respiratory failure or death [1,2,19]. The goal of COPD treatment is to reduce acute exacerbations of COPD [1,4,5]. One report showed that insulin sensitizers such as metformin may reduce the risk of severe COPD exacerbations [20]. Hyperglycemia tends to impair pulmonary function and increase the risk of acute COPD attacks [7,8,10]. Although insulin is the most effective glucose-lowering drug, our study showed that it might increase the risk of hospitalization for COPD. Insulin treatment has a direct effect on the lungs by inducing prostaglandin-mediated contraction of airway smooth muscle [21]. Insulin-like growth factor-1 (IGF-1) appears to play an important role in lung disease, but circulating IGF-1 has been reported to be associated with a lower risk of airway hyperresponsiveness [22]. Using insulin is likely to cause hypoglycemia and weight gain [8,23]. Prolonged use of insulin can lead to hyperinsulinemia, downregulation of insulin signaling and insulin resistance, which can increase glucose levels and oxidative stress. The downregulated insulin signaling leads to an imbalance in the anabolic activity of insulin, favoring protein synthesis and suppressing autophagy. The latter promotes cell senescence by inhibiting autophagic protein and lipid degradation and turnover [24]. All these factors can worsen lung function and exacerbate COPD.
Patients with COPD may exhibit alteration of the airway microbiome and impaired airway immune response, thus increasing susceptibility to bacterial infections [1,3,19].
Hyperglycemia tends to increase the permeability and spread of bacteria through the airways [8,25]. Diabetes mellitus is associated with a higher risk of bacterial pneumonia due to impaired cellular immunity [7,10]. We have previously shown that thiazolidinediones may increase the risk of developing bacterial pneumonia in people with COPD [26]. This study also found that insulin may increase the risk of bacterial pneumonia in people with T2D and COPD. The reason for this result is unclear. However, previous research has shown that hyperinsulinemia or insulin resistance can increase proinflammatory cytokines, predispose patients to inflammation, and impair immune function, thus increasing infection risk [23].
Hypoxia or respiratory failure can be a medical emergency in patients receiving COPD treatment [1] and often requires noninvasive ventilation to improve the patient's condition [2,4,19]. Intubation with invasive mechanical ventilation can be a life-saving procedure in patients with persistent hypercapnia [4]. This study showed that the use of insulin was associated with a higher risk of both NIPPV and IMV in people with COPD, and longer cumulative insulin use was associated with further increased risk of these outcomes. The use of insulin in patients with coexisting COPD and T2D may require attention to the possibility of respiratory failure.
Smoking is the main cause of lung cancer [1,[27][28][29]. Both COPD and T2D are risk factors for lung cancer [1,8,19]. A meta-analysis has shown that metformin may reduce the risk of lung cancer development [30]. Insulin can promote cell proliferation and mediate a mitogenic effect as binding to the insulin receptor. Insulin receptors can also bind to the insulin-like growth factor with potential oncogenic effects [31]. However, this finding has not been proved in large, long-term randomized controlled insulin trials [32]. Our study showed that insulin was associated with a non-significantly increased risk of lung cancer (aHR 2.11, 95%CI 0.99-4.49) in people with T2D and COPD. Insulin therapy in patients with COPD may need close monitoring for the occurrence of lung cancers.
Almost all studies on insulin therapy demonstrate the side effect of hypoglycemia [12][13][14]. Previous study has shown that about 37% of patients receiving insulin therapy developed hypoglycemia, and about 2.3% of patients with T2D had severe hypoglycemia annually [14]. Hypoglycemia can cause symptoms of cold sweating, palpitation, dyspnea, and general weakness. Kasirye et al. showed that hypoglycemia resulted in more severe acute exacerbation in patients with COPD [33]. Our research revealed that insulin use in patients with coexisting T2D and COPD could increase the risk of severe hypoglycemia. Dyspnea, weakness, acute exacerbation of COPD, and ventilation use could be attributed to severe hypoglycemia in patients on insulin therapy in this study. Elliot P Joslin stated, "Insulin is a remedy primarily for the wise and not for the foolish, be they patients or doctors . . . to use insulin successfully requires more brains" in 1928 [34]. Insulin therapy requires effective doctor-patient communication and careful monitoring for hypoglycemia.
COPD has been the third leading cause of death worldwide for years [1,4]. Diabetes mellitus can increase the severity of COPD and the risk of death [1,7,10]. Therefore, people with coexisting T2D and COPD are at high risk of death. Previous study has shown that metformin may have an effect on the risk of death in these patients [35]. Empagliflozin has been shown to have consistent benefits on cardiopulmonary and mortality outcomes in patients with T2D and COPD [36]. Our study showed that although insulin use might increase the risk of severe hypoglycemia, bacterial pneumonia, and ventilation use, it might not increase the risk of death in patients with T2D and COPD. Insulin studies tend to recruit patients with a longer duration of diabetes or more severe cases with a higher mortality rate. However, it is possible that this condition is not prominent in this study.
This study had several advantages. First, this was a nationwide population-based study. As 99% of Taiwan's 23 million people are covered by National Health Insurance, the patients enrolled in the study had less selection bias. Second, the study had a large sample size and an observational period of 18 years, which allowed for subgroup analysis and assessment of outcomes that took years to develop. Third, although this was only an observational study, no known randomized insulin studies have been conducted in patients with COPD recently, and the possibility of other randomized studies on insulin use in patients with COPD is low in the future [1,[7][8][9]. Therefore, this real-world data may fill some knowledge gaps in the clinical care of patients with T2D and COPD.
Our study had a number of limitations. First, national health administrative data lacked complete information on family history, individual lifestyle, smoking, and alcohol consumption, which could have influenced the findings of our study. In any case, we had well-matched baseline characteristics (all SMD values <0.1) and tried to reduce the difference between the study and control groups. Second, this administrative database lacked results of hemoglobin A1C levels, biochemical tests, pulmonary function tests, and imaging methods, which precluded accurate evaluation of T2D and COPD severity. However, we utilized DCSI scores, antidiabetics, and cardiovascular medications as representative factors to assess T2D status and the clinical records of moderate and severe COPD exacerbations to assess COPD severity. Third, patients on insulin therapy often exhibit a long duration of T2D, suboptimal glycemic control, and more comorbidities. Although we considered these factors, unknown residual confounders could bias the results. Fourth, the participants in this study were mainly of Taiwanese descent; therefore, the results may not be applicable to other racial groups. Finally, in a retrospective cohort study, there are always unobserved confounders. Therefore, there is a need for randomized control trials to confirm our findings.

Data Source and Population of the Study
We identified patients from the Taiwan's National Health Insurance Research Database (NHIRD), which was described in our previous study [25]. About 450 hospitals and over 10,000 clinics across Taiwan participate in the National Health Insurance program. The Health Insurance Authority conducts random inspections of all medical institutions annually to ensure the diagnostic accuracy and quality of care. The NHIRD has a link to the National Death Registry to confirm the diagnosis of death. The Declaration of Helsinki was followed for all methods used in this study. The China Medical University and Hospital Research Ethics Committee (CMUH104-REC2-115-CR4) approved this study. To prevent data leakage, detailed patient and provider information was encrypted and de-identified before release. The research ethics committee granted a waiver of informed consent.

Study Design
We identified persons with coexisting T2D and COPD from the NHIRD between 1 January 2000 and 31 December 2018, and followed them until 31 December 2019 (Figure 1). T2D and COPD were diagnosed with the ICD-9/10-CM codes for at least two outpatient visits or one admission in 1 year (Supplementary Table S1). Previous studies validated the ICD coding algorithms to define T2D and COPD with acceptable accuracy [37,38]. A moderate exacerbation of COPD was determined by the use of medication (systemic corticosteroids or antibiotics) and treatment in the outpatient claims; severe COPD exacerbations were defined as an acute attack treated at the emergency room (ER) or hospitalization [39]. Exclusion criteria were (1) age <18 or >80 years; (2) missing data of age or sex; (3) diagnosis of T1D, bacterial pneumonia, liver failure, or on dialysis; (4) hospitalization for COPD, bacterial pneumonia, ventilation use, or lung cancer, or death within 180 days of the index date to exclude latent morbidity; and (5) diagnosis of T2D or COPD prior to 1 January 2000 for the exclusion of prevalent disease.

Procedures
We defined the day of simultaneous diagnosis of T2D and COPD as the comorbid date ( Figure 3). In our database, there are 49,837 people from the diagnosis of T2D to T2D and COPD, 51,596 people from the diagnosis of COPD to COPD and T2D. Patients who received insulin treatment after the comorbid date were defined as insulin users, whereas those who never received insulin treatment during the study period were defined as insulin Pharmaceuticals 2023, 16, 643 9 of 12 non-users. The index date was defined as the first date of insulin use after the comorbid date. The index date of insulin non-user was considered to be the same period of time from the comorbid date to the index date of the insulin user. We assessed and matched the following variates that could interfere with the results: sex, age, obesity (the diagnoses of obesity, overweight, or severe obesity), smoking, comorbidities (hypertension, coronary artery disease, dyslipidemia, chronic kidney disease, atrial fibrillation, liver cirrhosis, and peripheral artery disease diagnosed one year before the index date), and medications (respiratory drugs, oral antidiabetics, antihypertensives, statins, and aspirin). Charlson Comorbidity Index (CCI) was assessed as a proxy for disease burden [40]. The Diabetes Complication Severity Index (DCSI) score [41] was used to assess complications of T2D; the occurrence of moderate or severe COPD exacerbations was used to assess COPD severity.

Procedures
We defined the day of simultaneous diagnosis of T2D and COPD as the comorbid date ( Figure 3). In our database, there are 49,837 people from the diagnosis of T2D to T2D and COPD, 51,596 people from the diagnosis of COPD to COPD and T2D. Patients who received insulin treatment after the comorbid date were defined as insulin users, whereas those who never received insulin treatment during the study period were defined as insulin non-users. The index date was defined as the first date of insulin use after the comorbid date. The index date of insulin non-user was considered to be the same period of time from the comorbid date to the index date of the insulin user. We assessed and matched the following variates that could interfere with the results: sex, age, obesity (the diagnoses of obesity, overweight, or severe obesity), smoking, comorbidities (hypertension, coronary artery disease, dyslipidemia, chronic kidney disease, atrial fibrillation, liver cirrhosis, and peripheral artery disease diagnosed one year before the index date), and medications (respiratory drugs, oral antidiabetics, antihypertensives, statins, and aspirin). Charlson Comorbidity Index (CCI) was assessed as a proxy for disease burden [40]. The Diabetes Complication Severity Index (DCSI) score [41] was used to assess complications of T2D; the occurrence of moderate or severe COPD exacerbations was used to assess COPD severity.

The Outcomes of Interest
The outcomes of interest in this study were all-cause mortality (the diagnosis of mortality was confirmed with the linkage to the National Death Registry), hospitalization for COPD, bacterial pneumonia, invasive mechanical ventilation (IMV), noninvasive positive pressure ventilation (NIPPV), lung cancers, and severe hypoglycemia (hypoglycemia requiring an emergency room visit or hospitalization). To calculate the outcomes, we followed patients until the date of death, respective outcomes, or at the end of follow-up on 31 December 2019, whichever came first.

Statistical Analyses
We used the Kelsey method and adopted the two-sided significance level of 95%, power of 80%, ratio of sample size between unexposed/exposed of 1, percentage of unexposed with outcome of 5%, and odds ratio of 1.5 to estimate the required exposed and unexposed sample size of 1690 and 1690, respectively [42]. Propensity score matching has been used for adjustment and comparability between insulin users and non-users [43]. We selected 32 clinically related variables as independent variates (including age, gender, comorbidities, DCSI and CCI scores, COPD exacerbation, and medications; Table 1) and insulin use as a dependent variable to estimate each patient's propensity score using non-

The Outcomes of Interest
The outcomes of interest in this study were all-cause mortality (the diagnosis of mortality was confirmed with the linkage to the National Death Registry), hospitalization for COPD, bacterial pneumonia, invasive mechanical ventilation (IMV), noninvasive positive pressure ventilation (NIPPV), lung cancers, and severe hypoglycemia (hypoglycemia requiring an emergency room visit or hospitalization). To calculate the outcomes, we followed patients until the date of death, respective outcomes, or at the end of follow-up on 31 December 2019, whichever came first.

Statistical Analyses
We used the Kelsey method and adopted the two-sided significance level of 95%, power of 80%, ratio of sample size between unexposed/exposed of 1, percentage of unexposed with outcome of 5%, and odds ratio of 1.5 to estimate the required exposed and unexposed sample size of 1690 and 1690, respectively [42]. Propensity score matching has been used for adjustment and comparability between insulin users and non-users [43]. We selected 32 clinically related variables as independent variates (including age, gender, comorbidities, DCSI and CCI scores, COPD exacerbation, and medications; Table 1) and insulin use as a dependent variable to estimate each patient's propensity score using nonparsimonious multivariable logistic regression. The nearest neighbor algorithm has been used for the construction of matched pairs, and a standardized mean difference (SMD) <0.1 was considered a negligible difference between the study group and the control groups. Using the propensity score matching method, a total of 2370 pairs of insulin users and non-users were identified.
Chi-squared tests were used to determine statistical differences in categorical variables, and Student's t-tests were used to determine statistical differences in continuous variables and between the study and control groups. To examine outcomes between insulin users and non-users, Cox proportional hazards models with multivariable adjustment were used. Results are presented as hazard ratios (HRs) and 95% confidence intervals (CIs). The Kaplan-Meier method and the log-rank test were used to compare the cumulative incidence of hospitalization for COPD, bacterial pneumonia, and invasive mechanical ventilation between insulin use and non-use. We have also compared the cumulative duration (<90, 90-179, >179 days) of insulin use and the risks of hospitalization for COPD, bacterial pneumonia, and invasive mechanical ventilation with insulin non-use. The cumulative duration of insulin was calculated by adding up from the index day to the end day of insulin use during the follow-up time.
Statistical significance was defined as a two-sided p-value of less than 0.05. Analyses of this study were performed using SAS (version 9.4; SAS Institute, Cary, NC, USA).

Conclusions
Patients with coexisting T2D and COPD require integrated, patient-centered, and multidisciplinary care. Pitifully, research on treatment strategies is scarce. Insulin may be indicated in diabetic patients with acute exacerbation of COPD to lower blood glucose. Our study has shown that patients with coexisting T2D and COPD requiring insulin therapy was associated with a higher risk of hypoglycemia, hospitalization for COPD, bacterial pneumonia, and ventilator use. However, we should be aware that this could be the result of the patients' intrinsic severe metabolic decompensation, which led to the previous administration of insulin. Insulin administration requires effective doctor-patient communication for careful monitoring of hypoglycemia, pneumonia, acute exacerbation of COPD, and respiratory failure with immediate and appropriate disposal.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/ph16050643/s1, Table S1: Diseases and related ICD-9-CM, ICD-10-CM codes. , and Tseng-Lien Lin Foundation, Taichung, Taiwan. These funding agencies did not influence the study design, data collection, and analysis; decision to publish; or preparation of the manuscript. The writing or preparation of this paper was not funded by any organization; the data analyses were not undertaken by individuals who are employees of funders, or any author who received funding from funders; and no writing support was provided by the funders. The corresponding authors had complete access to all data in the study and were responsible for the decision to submit this manuscript for publication. Institutional Review Board Statement: All methods of this study were performed in accordance with the Declaration of Helsinki. This research was approved by the Research Ethics Committee of China Medical University and Hospital (CMUH104-REC2-115-CR4).

Informed Consent Statement:
The detailed information of patients and health care providers was scrambled and de-identified before release to prevent data leaks. Informed consent was waived by the Research Ethics Committee.
Data Availability Statement: Data of this study are available from the National Health Insurance Research Database (NHIRD) published by Taiwan National Health Insurance (NHI) Administration. The data utilized in this study cannot be made available in the paper, the supplemental files, or in a public repository due to the "Personal Information Protection Act" executed by Taiwan government starting from 2012. Requests for data can be sent as a formal proposal to the NHIRD Office (https://dep.mohw. gov.tw/DOS/cp-2516-3591-113.html, accessed on 24 October 2022) or by email to stsung@mohw.gov.tw.