Time to Insulin Therapy and Severe Hypoglycemia in Korean Adults Initially Diagnosed with Type 2 Diabetes: A Nationwide Study
Article information
Abstract
Background
We examined the distribution of time to insulin therapy (TIT) post-diabetes diagnosis and the hazard of severe hypoglycemia (SH) according to TIT in Korean adults initially diagnosed with type 2 diabetes (T2D) and who progressed to insulin therapy.
Methods
Using data from the Korean National Health Insurance Service (2002 to 2018), we selected adult incident insulin users (initially diagnosed as T2D) who underwent health examinations between 2009 and 2012. The hazards of SH, recurrent SH, and problematic hypoglycemia were analyzed according to groups categorized using the TIT and clinical risk factors for SH (TIT ≥5 years with risk factors, TIT ≥5 years without risk factors, 3 ≤TIT <5 years, 1 ≤TIT <3 years, and TIT <1 year).
Results
Among 41,637 individuals, 14,840 (35.64%) and 10,587 (25.43%) initiated insulin therapy within <5 and <3 years postdiabetes diagnosis, respectively. During a median 6.53 years, 3,406 SH events occurred. Compared to individuals with TIT ≥5 years and no risk factor for SH, individuals with TIT <3 years had higher outcome hazards in a graded manner (adjusted hazard ratio [95% confidence intervals] for any SH: 1.117 [0.967 to 1.290] in those with 3 ≤TIT <5 years; 1.459 [1.284 to 1.657] in those with 1 ≤ TIT <3 years; and 1.515 [1.309 to 1.754] in those with TIT <1 year). This relationship was more pronounced in the non-obese subpopulation.
Conclusion
Among adults who progressed to insulin therapy after being diagnosed with T2D, a shorter TIT was not uncommon and may predict an increased risk of SH, particularly in non-obese patients.
INTRODUCTION
Severe hypoglycemia (SH) is an event that requires assistance from others for recovery due to the prominent neuroglycopenia [1,2]. In individuals with diabetes, SH is a lethal condition closely related to an increased risk of cardiovascular events, hospitalization for heart failure (hHF), and premature mortality [3-5]. Additionally, absolute endogenous insulin deficiency suggests compromised protection against declining plasma glucose levels and is thus strongly associated with SH, especially recurrent SH, in individuals with insulin-treated diabetes [6,7].
Previous studies in Western populations have shown that individuals who ultimately progress to absolute endogenous insulin deficiency, do so within 3 to 7 years post-diabetes diagnosis [8-10]. Furthermore, a population-based cohort study from the UK [9] showed that a rapid progression to insulin therapy was a potent predictor of future severe endogenous insulin deficiency, and 96% of individuals who eventually succumbed to absolute insulin deficiency initiated insulin treatment within 3 years of diabetes diagnosis. However, that analyses included only 586 participants with insulin-treated diabetes diagnosed after 30 years of age, and who had C-peptide measurements available. Thus, this may not represent the entire adult population with insulin-treated diabetes. Additionally, other ethnicities including East Asians, have not been included in previous studies conducted in Western countries.
In East Asia, including Korea, the incidence of typical autoimmune type 1 diabetes (T1D) is lower than in the other parts of the world [11,12]. However, other forms of diabetes requiring early insulin treatment, such as latent autoimmune diabetes in adults (LADA) or slowly progressive insulin-dependent diabetes mellitus (SPIDDM), have been reported to be relatively common [11,13]. Yet, the proportion of cases who require early insulin therapy among adults with diabetes ultimately progress to insulin treatment has not been systematically investigated in East Asia. Furthermore, little is known about the incidence of SH (including recurrent SH and problematic hypoglycemia) in the East Asian population with diabetes that require versus those that do not require early insulin intervention.
Therefore, to determine whether rapid progression to insulin therapy is closely related to the risk of SH and/or recurrent SH indicative of absolute endogenous insulin deficiency in an East Asian population with diabetes, we (1) examined the distribution of time to insulin therapy (TIT) post-diabetes diagnosis and (2) evaluated the hazards of SH according to TIT combined with other clinical risk factors among Korean adults initially diagnosed with type 2 diabetes (T2D) who progressed to insulin therapy. Additionally, recognizing the potential contribution of insulin resistance as another driver of early insulin therapy in individuals with obesity, we conducted a stratified analysis based on obesity status. We hypothesized that rapid progression to insulin therapy would be associated with an increased risk of SH, primarily due to a rapid decline in endogenous insulin production, rather than severe insulin resistance without significant endogenous insulin deficiency, which is typically seen with obesity.
METHODS
Data sources
We used data from the Korean National Health Insurance Service (KNHIS) from January 2002 to December 2018. The KNHIS, run by the Korean government, covers all Korean residents as a single nationwide insurer [14,15]. By assembling medical treatment, health screening records, and eligibility data from an existing database system, the KNHIS established a public database in 2012 [16]. From this database, demographics, monthly household income status, primary and secondary diagnoses expressed by the International Classification of Diseases10th Revision (ICD-10), prescriptions, procedures, dates of hospital visits and hospitalizations for all Korean residents, and date of death for the deceased are available in an anonymized form. Furthermore, the KNHIS actively operates a national health screening program and recommends standardized health examinations at least every 2 years through this program. These health examination results are also compiled into the database, and include smoking and alcohol history, physical activity, height, body weight, waist circumference (WC), blood pressure, and laboratory results including fasting plasma glucose (FPG) levels, lipid profiles, and estimated glomerular filtration rate (eGFR). Only hospitals accredited by the KNHIS carry out the standardized health examinations. Comprehensive information regarding this database can be found in prior publications [14-16].
The study protocol was approved by the Institutional Review Board (IRB) of Soongsil University (file number: SSU-202003- HR-201-01). The required for an informed consent was waived by the IRB, as the researchers were provided with de-identified, anonymous data from the KNHIS. This study was reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline for cohort study.
Study cohort, outcomes, and follow-up
This was a nationwide, longitudinal, population-based cohort study. Among adults (aged ≥20 years) with diabetes (initially diagnosed as T2D) who underwent ≥1 health examination between 2009 and 2012, we included incident insulin users who initiated insulin within 2 years after the examination (Supplemental Fig. S1). The timepoint of the first insulin prescription was considered baseline. As our study included only new insulin initiators at baseline, there was no period of insulin use prior to baseline, and TIT was defined as the time interval between the diagnosis of T2D (prior to or at baseline) and the timepoint of insulin initiation (baseline).
The presence of T2D prior to or at the time of health examination was defined as ≥1 claim per year for the prescription of anti-diabetes medication (under ICD-10 codes E11–14) or having a FPG level of ≥126mg/dL [11,17,18]. Based on the operational definition of continued insulin use in previous studies [19,20], to focus exclusively on persistent insulin users, we excluded individuals with ≤2 prescriptions of insulin in an outpatient setting during follow-up and those without claims for insulin prescription between 365 and 730 days from baseline. As such, we considered insulin use only as cases with at least three outpatient insulin prescriptions. Consequently, individuals who did not receive repeated insulin prescriptions in an outpatient setting and were only prescribed insulin on a short-term basis during hospitalization were excluded. Furthermore, as we considered only persistent insulin users—those with at least one insulin prescription between 365 and 730 days after the initial prescription—temporary users who did not maintain insulin use for more than 1 year were not included in the study. Furthermore, individuals who underwent total or partial pancreatectomy, received major organ transplantation, or were diagnosed with any malignancy at or before baseline, and those with missing data on ≥1 variable were excluded.
The endpoints were SH-related outcomes (any SH, recurrent SH, and problematic hypoglycemia). In accordance with the previous studies [3,4], SH was defined using the ICD-10 codes E16.x, E10.63, E11.63, E12.63, E13.63, or E14.63, from the inpatient or emergency room claims dataset. Any SH event was recorded as at least one SH event, whereas recurrent SH was defined as ≥2 SH episodes. Problematic hypoglycemia was defined as ≥2 events of SH per year [2]. From baseline, follow-up data was collected to the date of death, development of outcome, or December 31, 2018, whichever came first.
TIT and risk factors for SH
TIT was defined as the time interval between the diagnosis of diabetes (prior to or at baseline) and initiation of insulin therapy (baseline). TIT was categorized into ≥5, 3≤ TIT <5, 1≤ TIT <3, and <1 year to outline its distribution among a population representative of insulin initiators in South Korea. Since the KNHIS database is only available from January 2002, the exact date of diabetes diagnosis before 2002 could not be ascertained and categorizing TIT of ≥5 years in detail was not feasible. Participants with TIT of ≥5 years (who comprised a significant proportion) had heterogenous diabetes duration ranging from 5 years to several decades, and more varied degrees of complications and comorbidities as that groups include people with longstanding diabetes. Thus, individuals with TIT of ≥5 years were further subdivided into two groups according to the presence of ≥1 risk factor for SH. The risk factors were age of ≥65 years, presence of end-stage renal disease (ESRD), chronic kidney disease (CKD) at or before baseline, and previous SH at or before baseline [21]. Thus, a total of five groups were constructed according to the TIT and the presence of risk factors for SH (among those with ≥5 years of TIT). Other risk factors for SH such as the use of sulfonylurea, Charlson comorbidity index (CCI), and body mass index (BMI) were included as covariates in further multivariable analyses.
Measurements and definitions
Questionnaires were used to collect smoking history, alcohol consumption, and physical activity data. Blood tests including FPG and lipid profiles were done using venous samples drawn after an overnight fast. The eGFR was determined using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [22,23]. Definitions of low-income status, recent hospitalization, alcohol consumption status, regular exercise, hypertension, dyslipidemia, myocardial infarction (MI), stroke, atrial fibrillation (AF), hHF, CKD, ESRD, and BMI are summarized in Supplemental Table S1. The presence of obesity was defined as a BMI ≥25 kg/m2 [24]. The CCI was calculated using an established method [25] and previously summarized diagnostic codes [26]. Concurrent medication was examined based on prescriptions at baseline and in the 1 year before baseline. We also collected data on the use of sulfonylureas or meglitinide after insulin initiation, defined as prescription of these medications in the 1 year after the baseline.
Statistical analysis
Statistical analyses were conducted using SAS software version 9.3 (SAS Institute, Cary, NC, USA). Two-sided P values <0.05 were considered significant. The baseline characteristics of the study population are presented according to the five groups categorized by TIT and risk factors for SH. We presented continuous variables with normal distributions as mean±standard deviation, and variables with non-normal distributions as geometric means and 95% confidence intervals (CIs), and categorical variables as frequency and percentage.
The incidence rates of the outcomes were calculated as the number of incident cases divided by the total follow-up duration (person-years). The cumulative incidence rates of outcomes were compared among the five groups using Kaplan–Meier curves, and differences among groups were assessed using a log-rank test. Multivariable Cox regression analyses were conducted to estimate hazard ratios (HRs) and 95% CIs for the incidence rates of outcomes in the five groups. Model 1 was adjusted for age and sex; model 2 was adjusted for household income, smoking history, alcohol consumption, regular exercise, CCI, and recent hospitalization in addition to the variables included in model 1; model 3 was adjusted for major diagnostic codes at the time of first insulin prescription (codes for T1D or not), sulfonylurea use, and meglitinide use; model 4 was adjusted for the variables included in model 3, sulfonylurea use after insulin initiation and meglitinide use after insulin initiation; and model 5 was adjusted for BMI, FPG, eGFR, and SH before baseline in addition to the variables included in model 3. The definition of recurrent SH in this study was two or more SH events during follow-up. Individuals were categorized based on the occurrence of recurrent SH and problematic hypoglycemia (defined as ≥2 SH events per year), without analyzing the intervals between events. In the main time-to-event analysis for recurrent SH and problematic hypoglycemia, we only considered the time to the first SH event while we conducted additional sensitivity analysis to account for the recurrent nature of events. The proportional hazards assumption of the Cox models was assessed using Schoenfeld residuals.
Regarding the outcomes of recurrent SH and problematic hypoglycemia, we conducted additional sensitivity analyses to account for the recurrent nature of events. We employed the Andersen-Gill extension of the Cox proportional hazards model, which is designed for recurrent event data [27], and estimated the HRs of recurrent SH and problematic hypoglycemia in five groups. This approach is consistent with prior studies examining recurrent hypoglycemia episodes [28,29].
Given the differences in insulin resistance by the obesity status and the potential variance in mechanisms leading to the progression to insulin therapy, we reran the main analysis in subpopulations with or without obesity. This approach aimed to distinguish between individuals more likely to experience early insulin therapy due to endogenous insulin deficiency (non-obese) and those more likely to experience early insulin therapy due to insulin resistance (obese). In this analysis, model 5-1 was used, which is a variation of model 5 with BMI excluded as a potential confounding variable.
To exclude the influence of other risk factors and investigate the association between TIT and outcomes, we constructed restricted cubic splines for adjusted HRs and 95% CIs of outcomes according to TIT after excluding individuals with ≥1 risk factor for SH. To investigate the potential factors (other than TIT) associated with SH risk in individuals who progressed to insulin treatment within a relatively short time interval, we compared the characteristics of those with and without SH during follow-up only among those with TIT <3 years.
RESULTS
Baseline characteristics
A total of 41,637 incident insulin users (aged ≥20 years) were included (Supplemental Fig. S1). Among them, 14,840 (35.64%) initiated insulin therapy within <5 years after diabetes diagnosis, and 10,587 (25.43%) started insulin therapy within <3 years post-diabetes diagnosis (Fig. 1). Among those with TIT ≥5 years, 12,572 (46.92%) had ≥1 risk factor for SH. The baseline characteristics in each group are summarized in Table 1. Individuals who rapidly progressed to insulin therapy after a diabetes diagnosis were more likely to be current smokers, heavy drinkers, and have a history of recent hospitalization, although they tended to have a lower CCI. Furthermore, they were less likely to exercise regularly, use sulfonylureas or meglitinide at baseline, and use sulfonylureas after insulin initiation. Total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol levels tended to be higher in participants with a shorter TIT.
Distribution of time to insulin therapy (TIT) after being diagnosed with type 2 diabetes among incident insulin users.
Characteristics of the Participants according to Groups Categorized by Time to Insulin Therapy and Risk Factors for Severe Hypoglycemia
SH according to TIT
During a median 6.53 years (253,322.81 person-years), 3,406 SH events occurred in the entire cohort. The cumulative incidence of the outcome measures is presented according to the five groups categorized by the TIT and risk factors of SH using Kaplan–Meier curves (Fig. 2). The cumulative incidence of SH was the highest in individuals with TIT ≥5 years and with ≥1 risk factor for SH, and was the lowest in those with TIT ≥5 years and with no risk factors for SH. When individuals with TIT ≥5 years and no risk factor for SH were set as the reference, those with TIT ≥5 years and risk factors for SH had higher hazards of any SH, recurrent SH, and problematic hypoglycemia (HR in model 5: 1.520 [95% CI, 1.363 to 1.696] for any SH, 1.588 [95% CI, 1.360 to 1.855] for recurrent SH, and 1.777 [95% CI, 1.354 to 2.331] for problematic hypoglycemia) (Table 2). Compared to the reference, groups with shorter TIT (<3 years) exhibited progressively higher hazards of outcomes (HR for any SH in model 5: 1.117 [95% CI, 0.967 to 1.290] in those with 3≤ TIT <5 years, 1.459 [95% CI, 1.284 to 1.657] in those with 1≤ TIT <3 years, and 1.515 [95% CI, 1.309 to 1.754] in those with TIT <1 year). The HR for problematic hypoglycemia in model 5: 0.959 [95% CI, 0.649 to 1.416] in those with 3≤ TIT <5 years, 1.606 [95% CI, 1.172 to 2.200] in those with 1≤ TIT <3 years, and 1.888 [95% CI, 1.340 to 2.660] in those with TIT <1 year). This pattern of association was consistently observed in sensitivity analysis applying Andersen-Gill extension of the Cox proportional hazards model to account for the recurrent nature of events regarding the outcomes of recurrent SH and problematic hypoglycemia (Supplemental Table S2).
Cumulative incidence of (A) any severe hypoglycemia (SH), (B) recurrent SH, and (C) problematic hypoglycemia according to five groups categorized by the time to insulin therapy and risk factors of SH. The interval indicates the time interval from the diagnosis of diabetes mellitus to the initiation of insulin therapy (time to insulin therapy).
Hazard Ratios and 95% Confidence Intervals for the Incidence of Severe Hypoglycemia according to the Groups Categorized by Time to Insulin Therapy and Risk Factors for Severe Hypoglycemia
The pattern of graded inverse relationship was more pronounced in a subpopulation with BMI <25 kg/m2 (Supplemental Table S3), and much attenuated in subpopulation with BMI ≥25 kg/m2 (Supplemental Table S4). Rather, in cases where TIT was ≥3 years but <5 years, the hazards of recurrent SH and problematic hypoglycemia were lower compared to the reference (cases with TIT ≥5 years and no risk factor for SH) in a subpopulation with BMI ≥25 kg/m2. When restricted cubic splines were used with TIT as a continuous variable among individuals without any risk factors for SH (Fig. 3), until approximately 50 months of TIT, the hazards of any SH and problematic hypoglycemia showed a pattern of negative relationship with TIT, whereas the hazard of recurrent SH exhibited an inverted U-shaped relationship with a peak at 20 to 30 months of TIT and decreasing thereafter.
Hazard ratios (HRs) for (A) any severe hypoglycemia (SH), (B) recurrent SH, and (C) problematic hypoglycemia according to time to insulin therapy from the diagnosis of diabetes mellitus. During this analysis, individuals with one or more risk factors for SH were excluded. Curves represent HRs adjusted for age, sex, household income, smoking history, alcohol consumption, regular exercise, Charlson comorbidity index, recent hospitalization, major diagnostic codes at the time of first insulin prescription, sulfonylurea use, meglitinide use, body mass index, fasting plasma glucose level, estimated glomerular filtration rate, and SH before baseline. Solid lines indicate HRs and dashed lines indicate 95% confidence intervals (CIs) using restricted cubic spline regression.
Characteristics of individuals rapidly progressed to insulin therapy according to the development of SH
In individuals who progressed to insulin therapy within <3 years, baseline characteristics were compared between those with and without SH development during follow-up (Supplemental Table S5). Patients who experienced SH during followup were older, had higher CCI, higher prevalence of hypertension, MI, AF, hHF, and CKD, and lower eGFR than those who never experienced SH during follow-up. The proportions of people with T1D diagnostic codes (E10) as major diagnoses at baseline, those who suffered from SH before baseline, ≥3 oral anti-diabetes (OAD) medication users, and users of sulfonylurea at baseline and after insulin initiation were higher in those who developed SH during follow-up than the comparators. In addition, individuals with SH during follow-up exhibited lower BMI, WC, TC levels, and LDL-C concentration than those without SH during follow-up.
DISCUSSION
This nationwide longitudinal cohort study including 41,637 Korean adults initially diagnosed with T2D and who progressed to insulin treatment showed that more than one-third started insulin therapy within <5 years, and over one-fourth did so within <3 years after diabetes diagnosis. A shorter TIT (<3 years) was associated with a higher risk of SH, recurrent SH, and problematic hypoglycemia in a graded manner, especially in those without obesity.
Distribution of TIT in those who initiated basal insulin has been reported by several studies. In the EDITION 3 study, where 78% of the participants were Caucasian and 8.7% were of Asian/Oriental ethnicity, the mean TIT was 9.8±6.4 years [30]. Similarly, the mean TIT was 9.3±6.5 years in the First Basal Insulin Evaluation (FINE) study, which was a multinational, prospective, observational study of insulin-naïve people with T2D in Asia [31]. However, these studies did not specifically explore the proportion of those who required early initiation of insulin therapy in the general population.
To the best of our knowledge, this is the first population-based study in East Asians to examine the proportion of cases that required early initiation of insulin treatment among patients initially diagnosed as T2D and ultimately progressed to insulin therapy. In a previous study that determined the TIT from T2D diagnosis in primary care in Germany [32], median TIT (interquartile range [IQR]) in patients receiving incident insulin therapy was 4.70 years (IQR, 2.54 to 7.21) in 2010–2011, and 5.25 years (IQR, 3.07 to 8.08) in 2016–2017. The lowest quartile of TIT in their study [32] was comparable to our finding where more than a quarter (25.43%) of subjects initiated insulin therapy within 3 years of being diagnosed with T2D. According to the diabetes fact sheet in Korea 2021, only 7.5% of adults previously diagnosed as diabetes are treated with insulin [33]. Furthermore, significantly delayed initiation of insulin therapy has been documented in Korean patients with T2D uncontrolled with ≥2 OADs, indicating a substantial level of therapeutic inertia [34]. Therefore, it is unlikely that the high rate of transition to insulin treatment over a relatively short period of several years, as documented in our analysis, was due to medical professionals opting for early insulin therapy. Rather, it may be related to a significant proportion of patients being compelled to initiate early insulin therapy because of the relatively rapid progression of the disease and subsequent endogenous insulin deficiency (such as in those with genetic traits of reduced insulin secretory capacity, LADA, or SPIDDM). In East Asian countries (including Korea), although typical autoimmune T1D was reported to be uncommon compared to that in Western countries [11,12], other diabetes types requiring early insulin treatment are considered relatively common [11,13]. When the prevalence of LADA was compared in Caucasians and Asians using the same ascertainment criteria through a collaborative project between Italy and Korea, a marked similar prevalence of LADA was documented between these two countries with different ethnic populations [35]. This would be in line with the similarity in proportion of the study population progressed to early insulin therapy within 3 to 5 years after being diagnosed as T2D in our study and the previous study from Germany [32].
Although we could not clarify the exact underlying mechanism in this observational study, the association between shorter TIT and the risk of SH is not likely to originate from early insulin treatment itself or severe insulin resistance associated with obesity. Rather, it may occur due to an early insulin requirement associated with the progression to absolute endogenous insulin deficiency. In a study from the UK, rapid progression to insulin therapy within a few years was a strong predictor of severe endogenous insulin deficiency [9], which compromises protection against the development of hypoglycemia and is linked to the risk of recurrent SH [6,7]. In our study, in patients with obesity who are likely to have insulin resistance, a shorter TIT was not associated with an increased risk of SH. Instead, in cases where TIT was ≥3 years but <5 years, the hazards of recurrent SH and problematic hypoglycemia was lower than in those with TIT ≥5 years and no risk factors of SH. This suggests that early initiation of insulin therapy in obese patients with diabetes in this study was not necessarily due to a relatively rapid progression in endogenous insulin deficiency, compelling early initiation of insulin therapy. A randomized trial demonstrated that, compared with OADs, short-term intensive insulin therapy more effectively achieved glycemic control and improved beta-cell function in patients with new-onset T2D and severe hyperglycemia [36]. The results of this study might indicate that timely insulin therapy in individuals with obesity and diabetes, where the early requirement of insulin is less likely due to rapid progression to profound endogenous insulin deficiency but more likely due to severe insulin resistance, would not increase the risk of SH compared to those with delayed initiation of insulin therapy.
The clinical, anthropometric, and laboratory parameters varied in individuals with and without SH development who progressed rapidly to insulin treatment (TIT <3 years). Those who had SH during follow-up were leaner, suggesting a relationship with endogenous insulin deficiency. In addition, in those patients, factors such as age, comorbidities, renal function, previous SH events, and medication use including sulfonylureas varied. These findings are consistent with previous reports on the risk factors for SH among people with diabetes, which include old age, medication of sulfonylureas, renal dysfunction, a prior history of SH, multiple comorbidities [21,37,38]. These results indicate that clinical factors beyond the rapid progression to insulin therapy are also important to predict the risk of SH. Although the impact of other clinical factors was not the primary focus of this study and warrants further research, heightened vigilance for the incidence of SH would be necessary if clinical risk factors for SH were combined in patients with insulin-treated diabetes and a shorter TIT (e.g., <3 years). Additionally, modifiable factors, such as the concurrent use of sulfonylureas, should be adjusted in patients likely to be at high risk for SH.
Our study has several strengths. We included a large population of new insulin users (n=41,637) and analyzed a representative nationwide cohort database administered by the Korean government. As diverse variables including lifestyle, anthropometric, and laboratory data from a huge population are compiled in this KNHIS database, we could adjust for various covariates. Reporting distribution of TIT in a nationally representative cohort is clinically useful and can be crucial for decision-making by policymakers. For example, the fact that a substantial proportion of patients were compelled to initiate early insulin therapy due to the relatively rapid progression of the disease suggests that the number of people with T2D requiring insulin therapy would not be minimized, even with the future widespread use of oral glucagon-like peptide-1 analogues [39] and multi-agonists [40] in East Asia.
Nevertheless, the limitations of this study should be acknowledged. First, clarification of causal relationships was not possible because of the observational design. Second, due to the unavailability of data before 2002, TIT beyond 5 years could not be quantified in detail. Since we set the timepoint of the first insulin prescription as baseline and treated TIT and diabetes duration longer than 5 years as a single category, the impact of longstanding diabetes was not fully evaluated. Third, since our study population included only Korean adults initially diagnosed with T2D who started insulin therapy thereafter, extrapolation of our findings to other ethnicities or diagnoses should be done with caution. Fourth, as we selected the study population from health check-up examinees between 2009 and 2012, new insulin initiators who did not receive standardized health examinations may not be represented. The health check-up participation rate of the KNHIS in 2011–2012 was 72.6%–72.9% [15]. Fifth, despite our efforts to maximally adjust for measured potential confounders, unmeasured factors such as cognitive decline and prescribed insulin types might also have had effects. Lastly, the main time-to-event analysis for recurrent SH and problematic hypoglycemia only considered the time to the first SH event. However, the sensitivity analysis using the Andersen-Gill extension to account for the recurrent nature of the events demonstrated consistent results, and mitigated potential biases associated with the primary model.
In conclusion, in this real-world, nationwide, population-based study of Korean adults who progressed to insulin therapy after being diagnosed with T2D, more than one-fourth initiated insulin therapy within <3 years of diabetes diagnosis. This relatively shorter TIT (<3 years) was associated with an increased risk of SH, recurrent SH, and problematic hypoglycemia, particularly in those without obesity, indicating a possible early insulin requirement due to progression to profound endogenous insulin deficiency.
Supplementary Material
Supplemental Table S1.
Definitions of the Covariates
Supplemental Table S2.
Hazard Ratios and 95% Confidence Intervals for the Recurrent SH and Problematic Hypoglycemia according to the Groups Categorized by TIT and Risk Factors for SH, Sensitivity Analysis Applying Andersen-Gill Extension of the Cox Proportional Hazards Model to Account for the Recurrent Nature of Events
Supplemental Table S3.
Hazard Ratios and 95% Confidence Intervals for the Incidence of SH according to the Groups Categorized by TIT and Risk Factors for SH among Individuals with Body Mass Index <25 kg/m2
Supplemental Table S4.
Hazard Ratios and 95% Confidence Intervals for the Incidence of SH according to the Groups Categorized by the TIT and Risk Factors of SH among Individuals with Body Mass Index ≥25 kg/m2
Supplemental Table S5.
Baseline Characteristics of Participants with or without SH during Follow-up among Those with TIT <3 Years
Supplemental Fig. S1.
Flow diagram of the study population.
Notes
CONFLICTS OF INTEREST
No potential conflict of interest relevant to this article was reported.
AUTHOR CONTRIBUTIONS
Conception or design: Y.B.L., K.H., S.H.P., S.M.J. Acquisition, analysis, or interpretation of data: K.H., B.K., K.Y.H., G.K. Drafting the work or revising: Y.B.L., J.H.K. S.M.J. Final approval of the manuscript: Y.B.L., K.H., B.K., S.H.P., K.Y.H., G.K., J.H.K. S.M.J.
Acknowledgements
This work was supported by a grant from the Korea Medical Device Development Fund funded by Ministry of Science and ICT, Ministry of Trade, Industry and Energy, Ministry of Health & Welfare, Ministry of Food and Drug Safety (Grant Number: RS-2020-KD000056). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. We used data from the Korean National Health Insurance Service (KNHIS). We used the National Health Information Database constructed by the KNHIS, and the study results do not necessarily represent the opinions of the KNHIS.