1. Study design

This study was a secondary analysis of the data from the Ansan cohort of the KoGES, which was a community-based cohort study conducted by the Korea Disease Control and Prevention Agency.

For this study, the application documents for data sharing were submitted through online data sharing service of the division of epidemiology and health index, the National Institute of Health, the Korea Disease Control and Prevention Agency. The epidemiologic data of the Ansan cohort of the KoGES from baseline survey (2001~2002) to fifth follow-up visit (2011~2012) were downloaded after approval notification.

2. Study participants

The Ansan cohort represented small/medium-sized cities in Korea and comprised 5,012 adult males and females aged between 40 and 69 years. The baseline survey of the Ansan cohort was completed in 2001 and 2002, and follow-up surveys are conducted every other year. Details of the KoGES and Ansan cohort have been described previously [18, 21, 22, 23]. In brief, this study used a two-stage cluster sampling for subject enrollment at the Ansan site, based on the local telephone directory information and information on demographic characteristics from the 2000 census. Altogether, 10,957 eligible individuals were identified by telephone contact; 5,012 subjects participated in the baseline survey, and 3,052 subjects remained until the fifth follow-up survey (60.9% follow-up rate). The mean age ± standard deviation (SD) was 49.1 ± 7.9 for the 5,012 subjects at baseline and 58.3 ± 7.3 for the 3,052 subjects at the fifth visit. All subjects were surveyed for demographic and medical history, family history of disease, and lifestyle habits (including smoking, alcohol use, physical activity, and dietary intake) using an interview-based questionnaire and they underwent comprehensive health check-ups and provided bio-specimens for assays.

Subjects were included in this study upon meeting the following criteria: people with type 2 DM, without cataract at the baseline survey, and completed at least one follow-up survey. The sample size was calculated using the cohort study design of the Open Source Epidemiologic Statistics for Public Health [24], an online sample counting program. Based on the results of the duration of DM (< 10 years versus ≥ 10 years) in Becker et al. [5], a significance level (α) of .05, power (1 - β) = .85, percentage of unexposed with outcome= 20.5%, and percentage of exposed with outcome = 42.0% result in a minimum required sample size of 256 (Kelsey method) to 311 (Fleiss with correction method). Among 5,012 people who participated in the baseline survey, 329 who met the inclusion criteria were included in the analysis. Eventually, cataract occurred in 63 (19.1%) patients during the 10-year follow-up period (Figure 1).

Figure 1
Flow of participants through the study.
DM = Diabetes mellitus; KoGES = Korean Genome and Epidemiology Study.

• Download Figure

3. Definition and variables

Following previous studies [21, 22, 23], DM subjects were defined as those who met at least one of the following four criteria: 1) self-reporting of DM, 2) undergoing insulin or oral hypoglycemic agent treatment, 3) with fasting plasma glucose (FPG) ≥ 126 mg/dL or postprandial 2-hour glucose (PP2) ≥ 200 mg/dL after a 75-g oral glucose tolerance test (OGTT), and 4) with HbA1c ≥ 6.5%.

Cataract subjects were classified according to their response to the questions in medical information section about the experience and timing of cataract diagnosis after DM diagnosis during the third to fifth follow-up surveys. The question was, “Have you been diagnosed with cataract in relation to DM after DM diagnosis?” Those who responded “No, I have never been diagnosed” to the question were classified as having “no cataract” while the rest were classified as having “cataract”. The timing of cataract diagnosis was coded as “before 2001”, “2001~2002”, “2003~2004”, “2005~2006”, “2007~2008”, “2009~2010”, and “2011~2012”. Meanwhile, if they responded “before 2001” or “2001~2002” to the question, did not respond to the question, or did not participate in the follow-up survey, they were classified as “having history of cataract at baseline”, “unknown cataract” or “no survey”, respectively, and all were excluded from the study.

The explanatory variables, selected based on previous studies on cataract-related factors [5, 16, 17, 25, 26, 27], were socio-demographics, health behaviors, and disease-related characteristics of the subjects measured at the baseline survey. For all characteristic variables, KoGES data were used as it was or re-categorized. They included sex, age, education, family income, smoking status, total smoking amount (pack-year), status of alcohol use, total alcohol intake (g/day), total physical activity, BMI, age at the initial diagnosis of DM, use of oral hypoglycemic agents, insulin treatment, FPG, PP2, HbA1c, fasting plasma insulin, comorbidities (hypertension, hyperlipidemia, asthma, ischemic heart disease, congestive heart failure, myocardial infarction, chronic obstructive pulmonary disease), systolic blood pressure, diastolic blood pressure, taking of lipid-lowering agents, steroid use, and lipids (total cholesterol, low-density lipoprotein cholesterol [LDLC], high-density lipoprotein cholesterol [HDLC], and triglycerides). Some of these variables were converted and used according to the purpose of this study. Regarding education, the subjects were divided into two categories: high school graduate or not. Family income per month was divided into two categories: below or above three million KRW which was the closest to the average family income in 2003 [28]. For smoking status and status of alcohol use, the subjects were divided into two categories: current smoking or not and current alcohol use or not, respectively. Physical activity was measured using the metabolic equivalent of task. BMI was calculated using height and weight which were measured using standard methods (weight divided by height squared, kg/m²). The duration of DM was calculated by subtracting participants’ age at the initial diagnosis of DM from their age at the baseline survey, and those who met only the laboratory criterion of DM were classified into the “not applicable” group. FPG, fasting plasma insulin, and triglycerides were measured in the central laboratory by collecting blood after fasting for 12 hours. PP2 was checked by measuring plasma glucose at 2 hours by a 75 g OGTT. LDLC was calculated using three lipid data: LDLC = total cholesterol–[(triglyceride/5) + HDLC] [29]. Insulin resistance was calculated using the homeostasis model assessment of insulin resistance (HOMA-IR). The equation for HOMA-IR was “HOMA-IR = [fasting insulin (µIU) × fasting blood sugar (mg/dL)] / 405” [22, 30, 31]. Hypertension was defined as subjects who had left or right systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg, or subjects who were taking anti-hypertensive agents [30, 32]. Meanwhile, steroid use and comorbidities except hypertension and hyperlipidemia were finally excluded from the explanatory variables because there was no or small subjects (1~6 cases) who took steroid or had those comorbidities.

4. Data analysis

Data were analyzed using IBM SPSS Statistics for Windows (version 23.0; IBM Corp., Armonk, NY, USA) and Open Source Epidemiologic Statistics for Public Health [24], and the significance level of the two-tailed statistical test was .05. The 10-year incidence of cataract was calculated as the incidence proportion (%) and incidence rate per 1,000 person-years with their respective 95% CIs. Person-years were calculated from the age at the baseline survey until the age at cataract development or the last survey.

Hazard ratios (HRs) and 95% CIs were calculated using a simple Cox proportional hazards model to compare the survival patterns according to the study subjects’ baseline characteristics. A log minus log plot was used to confirm whether each variable met the proportional hazard assumption [33]. The survival time was defined as the period from the baseline survey to cataract occurrence, and calculated in units of year by subtracting the age at the time of the baseline survey from the age at the time of cataract occurrence. A multivariate Cox proportional hazards model was used to identify the factors related to cataract development with significant variables from the simple model as explanatory variables, except for the duration of DM. For the duration of DM, some data (163 in total) corresponding to the “not applicable” group were missing. Thus, demographic variables (sex, age, education, and family income), health behaviors (current smoking and current alcohol use), and disease-related variables (HbA1c) were entered into the model in a series of steps, the HRs and their 95% CIs and -2log likelihood values were calculated. The analysis results, including the duration of DM in the Cox model, are presented as a supplement. The correlation coefficient (Spearman’s rho) among the variables ranged from .02 to .48, which satisfied the assumption of multicollinearity.

Meanwhile, about half of the subjects with DM at baseline were excluded for “unknown cataract” or “no survey” on cataract history; a sensitivity analysis was performed to check the robustness of the findings. Specifically, we compared the baseline characteristics of those who were included (n = 329) and excluded (n = 305) in this study using the χ² test, Fisher’s exact test, t-test, or Mann–Whitney U test.

5. Ethical consideration

This study was conducted after receiving approval for exempt review (PNU IRB No. 2018_24_HR) from the Institutional Review Board (IRB) of Pusan National University. Our data were received from the department in charge of KoGES in the KDCA after submission of the IRB approval letter and study proposal to the Department, and all data were anonymized.

1. Cataract incidence

Among the 329 subjects, 63 developed cataracts, with an incidence proportion of 19.1% (Table 1). The incidence rate per 1,000 person-years was 22.2 (95% CI = 17.0~28.4), 17.5 (95% CI = 11.9~24.8) among males and 29.9 (95% CI = 20.5~42.2) among females. The mean age at the incidence of cataract was 63.5 years. The mean elapsed time from the baseline survey until cataract development was 5.4 years. For females, the elapsed time before contracting cataract (5.9 years) was significantly longer than for males (4.8 years; Z = -2.17, p = .030). Meanwhile, the mean elapsed time from DM diagnosis to cataract development was 13.3 years, and there was no difference by sex.

Table 1
Cataract Incidence-Related Characteristics of Study Subjects (N = 63)

• Download Table

2. Cataract incidence according to the subjects’ characteristics at the baseline data

Table 2 shows that subjects with and without cataract were significantly different in their characteristics at baseline: females (p = .034), mean age (p < .001), education (p = .003), family income (p = .004), current smoking (p = .043), current alcohol use (p = .009), duration of DM (p < .001), and HbA1c (p = .016). In the cataract group at baseline, there were more females; more current smoking and fewer current alcohol use; the group members were older, less educated, and had less family income, longer duration of DM, and higher HbA1c than the no-cataract group members.

Table 2
Comparison of Baseline Characteristics of Study Subjects with or without Cataract Development (N = 329)

• Download Table

3. Factors related to cataract incidence

Among the seven variables, except for the duration of DM, which were significant in the simple model, age (HR = 1.09, 95% CI = 1.05~1.13) and HbA1c (HR = 1.21, 95% CI = 1.07~1.37) were independently associated with the development of cataracts in multiple models (Table 3). Meanwhile, in the additional analysis that included the duration of DM in the multiple model, age (HR = 1.08, 95% CI = 1.03~1.13) and the duration of DM (HR = 1.05, 95% CI = 1.00~1.09) were significantly related to the development of cataracts (Supplementary Table 1).

Table 3
Factors Related to the Cataract Development (N = 329)

• Download Table

4. Characteristics of the subjects included and excluded in the study

According to the sensitivity analysis, the characteristics of the subjects included in the study and those excluded were significantly different for several variables. The group included in the study had fewer females (p = .036), and compared with the excluded group, its members were less educated (p = .005), had hypertension (p = .010), lower HDLC (p = .013) but higher BMI (p = .025), FPG (p < .001), PP2 (p < .001), HbA1c (p < .001), and HOMA-IR (p = .007; Supplementary Table 2).

1. Strength and limitations

To our knowledge, this study is one of few studies that used a longitudinal study design to investigate cataract incidence and its predictive factors among people with DM. Considering that in South Korea, only about 60% of males and 68% of females aged ≥ 30 years with DM are aware of their DM status [41], the use of laboratory results as inclusion criteria may be helpful in minimizing the exclusion of DM patients. However, careful interpretation of the results is necessary because of the following limitations. First, because this study relied on self-reported cataract data, there may be problems with accuracy, as we could not validate the cataract diagnosis and underestimation and recall bias, compared with clinical tests [42]. Therefore, further studies are recommended to use objective outcome data. Second, the cataract type could not be considered in this study because of lack of data. Srinivasan et al. [16] analyzed subjects of the Sankara Nethralaya Diabetic Retinopathy Epidemiology and Molecular Genetic Study and found that the incidence of cataract greatly differs between cataract types, that is, 7.3% for posterior subcapsular cataract and 70% for nuclear opalescence. This finding highlights the importance of further investigation into the incidence rate according to cataract type. Third, the time until cataract occurred, required for calculating cataract density, was measured using the difference in age between the time of the baseline survey and that of cataract incidence, entered in the KoGES. This is because it is difficult to identify the actual time of cataract incidence since KoGES is only conducted every other year. However, this method raises the possibility of recall bias because the age when cataract occurred was self-reported. Fourth, as the study subjects were those with prevalent type 2 DM, the duration of DM varied at the baseline survey. Accordingly, some of their characteristics at baseline might differ from those at the actual time of their DM development, although the duration of DM was used as an explanatory variable in the data analysis to control for the effect of DM progression. Further longitudinal studies with DM incident cases to identify factors related to cataract development are suggested. Fifth, the duration of DM was missing in about half of the participants because DM subjects were identified using self-report of DM diagnosis and medication as well as laboratory results based on previous studies [21, 22, 23]; this may lead to substantial loss of statistical power in multivariate regression analysis. Sixth, although compliance to treatment and follow-up in people with DM could affect glycemic control and in turn the occurrence of cataract, this dynamic could not be assessed here because of lack of data in the KoGES. Lastly, the generalizability of the findings to other population groups, except for the study subjects, is limited. A substantial proportion of individuals with DM at baseline (n = 305 [48%]) was excluded because of unknown, or no survey on, cataract history; the small sample size may decrease the representativeness of the study results. There were significant differences between included and excluded individuals on several characteristics, which may have increased selection bias. Therefore, caution should be exercised when interpreting and applying the results of this study to all people with DM living in Ansan as well as other small and medium-sized cities or rural areas in South Korea. The results of this study should be interpreted as being limited to the subjects of this study.

2. Clinical implications

Cataracts, one of the most common causes of blindness in adults, lead to a decrease in quality of life and an increase in economic burden. Considering that people with DM have twice as much risk of cataract development than those without DM, nurses working at hospitals or healthcare facilities in the community should educate people with DM about the major risk factors for cataract development based on the findings of this study, and recommend regular eye examinations so that cataracts can be identified as early as possible. Nursing researchers can help monitor continuously the magnitude of cataract risk in people with DM by constructing and validating a risk factor prediction model using the risk factors identified in this study. Finally, national health policy-makers can consider eye examinations in the health check-up items for those who have a high risk of cataract among people with DM or suspected to be in high glycemic condition.