Study design
This was a retrospective, cross-sectional study using a commercially available Japanese healthcare database (MinaCare Co. Ltd., Tokyo, Japan). The main aim of the study was to examine BP, LDL-c and HbA1c values from health checkup data from a cross-sectional viewpoint, where a single time point for each individual was selected for analysis. Smoking status, body mass index (BMI), and waist circumference were also evaluated, because these three measures are also known risk factors for CVDs. MinaCare data were then compared with the latest (at the time of analysis) publicly available summary results from MHLW-SH (2010) [8] and MHLW-H&N (2011) [7].
MinaCare database
MinaCare Co. Ltd. manages healthcare data provided by employment-based health insurance and offers health risk reduction plans to corporate employees diagnosed with obesity or those who have a high risk for any lifestyle-related diseases. The MinaCare database includes regularly updated data of checkups and medical and pharmaceutical claims since 2010 (Figure 1). The population covered by the database includes working individuals and their dependent family members, and the database covers a wide range of age groups. Employment-based insurance covers large-scale, nation-wide retailers, manufacturers, and those in food, information, transportation and energy industries. Therefore, the MinaCare database includes health information of individuals with minimal geographic or occupational bias, but it has the limitation that individuals in the primary industries such as agriculture, fishery and forestry, and those who are self-employed are not included. Health checkup data include information on subjects’ demographics, smoking status, vital signs, clinical laboratory test data, and administrative information. Data of medical and pharmaceutical claims from approximately 1.1 million subjects that are 3%-4% of the total insured individuals in Japan, and health checkup data from approximately 338000 subjects were collected into the MinaCare database as of November 2013. Considering the nature of MinaCare Co. Ltd, which manages healthcare data provided by employment-based health insurance, and the fact that insurers must offer subscribers an annual health checkup, it seems obvious that a high percentage of the target population undergoes annual checkups.
Anonymization of subjects
This study used subject-level electronic health related databases that protected the identity of individuals. MinaCare is allowed to use such anonymized data under the data transfer contract with its client health insurers. We referred the “Ethical Guidelines for Epidemiological Research” laid down by the Japanese Government [11], although this study is outside the scope because the data is protected by anonymization.
Data extraction and management of irregular data
All health checkup data included in the MinaCare database for the fiscal years from 2010 to 2012 were extracted. Health checkup examinations are normally conducted annually, thus this dataset included subjects with examination years that ranged from 1 up to a maximum of 3 years. In this study, all subjects who had at least one set of health checkup data for the fiscal year 2011 were selected (a total of 232515 subjects). For a few subjects who underwent multiple examinations in the same year, earliest examination data were used. It was noted that 2011 was a representative cross-sectional year for the range of fiscal years from 2010 to 2012 and was also a single year that was most comparable with the timing of the two national surveys.
Missing values were not imputed and were excluded from analysis, unless otherwise noted. Non-numeric or impossible values for numeric variables were handled as missing values and were also excluded from the summary. The proportion and pattern of missing values were monitored to consider any implications in the interpretation of the results.
Study variables
General information (subject ID, year of examination, birth year/month, and date of examination), sex, body size measurements (height, weight, waist circumference, BMI, obesity score), smoking status, vital signs (average BP), and clinical laboratory tests [fasting blood glucose (FBG), HbA1c, urine sugar, uric protein, total cholesterol, triglyceride (TG), high density lipoprotein cholesterol (HDL-c), and LDL-c] data were extracted from the database. In addition, the age at the time of examination was computed from the birth year/month by imputing the first day of the month for the birth date. Smoking status was based on a subject questionnaire regarding the subject’s current smoking status.
In terms of HbA1c values, the values of the Japan Diabetes Society (JDS) were used in Japan in 2011 for health checkups. HbA1c (JDS) ≥6.1%, comparable with HbA1c [National glycohemoglobin standardization program (NGSP)] ≥6.5%, was the diagnostic criterion for diabetes mellitus based on the JDS guideline [6, 12].
Study endpoints
The primary endpoints of this study were the distribution of demographic variables, the distribution of BP [systolic BP (SBP) and diastolic BP (DBP)], LDL-c, and HbA1c values, and the proportion of subjects who achieved the predefined BP, LDL-c, and HbA1c cutoff levels. These endpoints were examined in various subgroups of interest.
The secondary endpoints of this study included the distributions of HDL-c, TG and FBG values and the proportion of subjects who achieved the respective predefined cutoff levels and the proportion of subjects within each cardiovascular and cerebrovascular risk level. These endpoints were examined in various subgroups of interest.
Cutoff levels for BP (SBP and DBP), lipid parameters (LDL-c, HDL-c, and TG), HbA1c and FBG were based on the JSH 2009 guideline [4], JAS 2012 guideline [5] and JDS 2013 guideline [6], respectively.
Age (at examination), sex, height, weight, BMI, waist circumference, and smoking status were considered as covariates. Subgroups defined by the levels of the covariates were examined.
Statistical method
This analysis was primarily based on descriptive statistical methods, and formal statistical inference was not used. For continuous data, summary statistics such as mean, median, minimum, maximum, and standard deviation (SD) were used. For binary and categorical data, summary statistics such as numbers and proportions (percentages) were used. Graphical presentations were used to describe data. To enhance visual comparisons, graphical presentations of means and proportions of various endpoints were made without standard errors or confidence intervals. Because the number of subjects (n) in each sex and age category was generally very large (as seen in Figures 2 and 3), this simplification did not affect the appropriate interpretation of data. However, special notes on variability were made when indicated.
The Structured Query Language was used to extract data from the MinaCare database. SAS version 9.2 was used to derive datasets for statistical analyses. R (version 2.15.2) and MS EXCEL 2007 were used to create graphs.
Existing data sources used for the comparison with MinaCare database
The summary level results of MHLW-SH data (2010) and MHLW-H&N data (2011) were obtained from existing reports open to the public on the MHLW homepage [7, 8]. The number of reported subjects (sex tabulation) was 22232094 (12228976 males and 10003118 females) in MHLW-SH data (2010) and 6914 (3159 males and 3755 females) in MHLW-H&N data (2011).
Classification of adult BP values by the JSH 2009 guideline
Subjects were classified according to their BP levels based on the JSH 2009 guideline [4]. In this guideline, the following six categories were defined based on combination of SBP and DBP values: ideal, normal, normal (high), HT class I, HT class II, and HT class III.