Study design and population

We conducted a nationwide cohort study, known as the Metabolic Syndrome And Comprehensive lifesTyle Intervention study On Nationwide database in Japan (MetS ACTION-J) with retrospectively collected data from the National Database of Health Insurance Claims and Specific Health Checkups of Japan (NDB). The national data captured examination records and laboratory data to diagnose MetS. We obtained anonymized data regarding subjects who underwent a screening program between fiscal year 2008 and fiscal year 2011 from the Ministry of Health, Labour and Welfare (MHLW). We have created a database for this study (S1 Appendix). The ethics committees of the University of Kyoto School of Medicine and National Cerebral and Cardiovascular Center approved the study. Written informed consent was not obtained due to the retrospective design.

We considered all subjects aged 40–74 years who completed the screening visits in fiscal year 2008. We excluded subjects who were receiving medications for hypertension or dyslipidaemia or diabetes, those who meet the diabetes criteria (fasting blood glucose [FBG] ≥126 and/or hemoglobin A1c [HbA1c] ≥6.5%), those without data required for diagnosing metabolic syndrome (MetS), and those who did not have the specific health checkup in fiscal year 2011.

Screening and lifestyle interventions

All subjects received health checkups, including questionnaires, physical examinations, and measurements of height and weight, waist circumference (WC), blood pressure, serum cholesterol, blood glucose, and/or HbA1c. All exams were conducted under an article prescribed by an ordinance of the MHLW (S2 and S3 Appendices). Smoking status was determined by self-reported questionnaires. Exams, including anthropometry measurements, blood pressure, and blood samples, were carried out after at least a 10-hour fast, and the subjects were instructed to perform no physical activities of moderate or high intensity and to ingest no caffeine and alcohol the day before the test. WC was measured at the umbilical level in a standing position using a tape measure after normal expiration (S2 Appendix). If the umbilical level was displaced downward due to accumulation of abdominal fat, WC was measured at the midpoint between the superior border of the iliac crest and the inferior margin of the twelfth rib. Blood pressure was measured twice after at least 5 min rest with the participant seated. Regardless of the presence/absence of risk, information and advice on CVD prevention were given to all subjects simultaneously with notification of the checkup results annually or more frequently. Specific health promotion guidance was provided to at-risk individuals. Individuals at risk were defined as those with a WC ≥85 cm for men and ≥90 cm for women [8] and/or body mass index (BMI) ≥25 kg/m² who satisfy at least one of the following requirements: (1) elevated blood pressure (systolic blood pressure [SBP] ≥130 mmHg or diastolic blood pressure [DBP] ≥85 mmHg), (2) dyslipidaemia (triglycerides [TG] ≥150 mg/dl or high-density lipoprotein [HDL]-cholesterol <40 mg/dl), and/or (3) impaired glucose tolerance (FBG ≥100 mg/dl or HbA1c ≥5.6%). After subjects with risks received health checkup results by mail or other means, they reserved counselling sessions themselves. Health promotion guidance was provided via interview by healthcare providers (physician, health nurse, or managerial dietician) with or without additional continuous support for 3 months or longer in response to individual risks (Table 1 and S3 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Nationwide preventive program: Screening and lifestyle intervention features.

https://doi.org/10.1371/journal.pone.0190862.t001

Statistical analysis

We evaluated WC and BMI reduction, and reversal of MetS using logistic regression models and changes in cardiometabolic risk factors using linear regression models to investigate associations between lifestyle intervention and cardiometabolic risk factors. Since previous studies have shown that a loss of 5–10% of initial weight is associated with significant reductions in cardiovascular risks, a 5% reduction was considered clinically meaningful [9]. Reversal of MetS was defined as normalization for subjects with MetS [8] and risk reduction for subjects with pre-MetS. We performed regression analysis with coefficients expressed per one standard deviation (SD) to compare intervention effect size directly among cardiometabolic risk factors. Covariates included age, sex, and smoking status as categorical variables, and BMI (except the WC model) and baseline individual component level as continuous variables. In addition to adjusting for covariates, we performed rigorous adjustment for baseline variables using propensity-score matching. The propensity score was calculated from a multivariate probit regression model in the whole cohort that included demographic characteristics and cardiometabolic risk factors. Matching was performed using a 1:1 matching protocol within a calliper of 0.01 SD of the propensity score probit.

We performed IV analysis to provide estimates that would remain unbiased even if important confounding variables were not measured [10]. We used the facility level ratio of lifestyle intervention, in which the facility intervention ratio was defined as the number of intervention participants divided by the total number of candidates, as the instrument. Similar instruments have previously been reported [11]. In this study, it is unlikely that the facility ratios of lifestyle intervention would be associated with clinical improvements in any way other than the participants of intervention. To confirm that the ratio of lifestyle intervention was not a weak instrument, we used a partial F test and found sufficiently large F statistics. Patients seen at institutions with an intervention rate less than 0.1% and over 60% were excluded from the IV analysis, since lifestyle intervention choice may have been decided less by facility preference and therefore be subject to confounding by unmeasured subject-level covariates [12]. We evaluated the balance of measured covariates across levels of the IV to provide additional information to assess its validity (S2 Table). Using two-stage linear regression for the calculation of the coefficients, we estimated IV-adjusted changes of cardiometabolic factors with the IV being the facility intervention rate, using the STATA procedure IVREG. Covariates include geographic location and clinical characteristics, including urban or rural (S4 Appendix), age, sex, smoking status, BMI (except the WC), and baseline individual component level. The analyses were conducted using STATA, 13.1 (STATA Corp., College Station, Texas) and SAS 9.2 (SAS Institute, Cary, North Carolina).

Study population

We included subjects (n = 19,969,722) aged 40–74 years who completed the screening visits in the first year (from April 2008 to March 2009). Among this population, we excluded 15,599,680 subjects who were receiving medications for hypertension or dyslipidemia or diabetes (n = 5,051,629), those who meet diabetes criteria (n = 417,450), those without appropriate data required for diagnosing MetS (n = 1,793,327), those with data error (n = 879,889), and those who did not have the specific health checkup in fiscal year 2011 (n = 7,457,385). Of 4,370,042 subjects who were eligible for analyses, 3,350,354 (76.7%) were eligible for healthcare guidance (information supply only), and 1,019,688 (23.3%) were eligible for the interventional program. Totally, 111,779 subjects attended the health guidance program (participants), and 907,909 did not (non-participants) (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study flow chart.

FY, fiscal year.

https://doi.org/10.1371/journal.pone.0190862.g001

At baseline, the mean (SD) WC and BMI were 91.2 (5.9) cm and 26.2 (2.5) kg/m² for participants and 91.2 (5.7) cm and 26.1 (2.4) kg/m² for non-participants, respectively. Non-participants had slightly worse blood pressures and lipid profiles than participants (Table 2). After propensity-score matching was performed for the entire population, there were 75,777 matched pairs. The absolute standardized differences were <10% for all variables entered into the propensity score, indicating adequate matches [13].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Baseline characteristics.

https://doi.org/10.1371/journal.pone.0190862.t002

Effectiveness of nationwide lifestyle intervention

Clinically relevant reductions, i.e., ≥5% at year 3, were achieved in a significantly higher percentage of participants compared to non-participants (WC, 21.4% vs 16.1% and BMI, 17.6% vs 13.6%; p<0.001, each; Fig 2). Advanced weight reductions (≥10%) were also observed in a significantly higher percentage of participants. Both abdominal and overall obesity, measured by categorical weight reductions, improved significantly in participants compared to non-participants. Participants who received lifestyle intervention had significantly improvements in MetS, as compared with non-participants (reversal of MetS: 47.0% vs. 41.5%, p<0.001). Analysis in the matched cohort showed similar results (S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Categorical waist and weight reduction.

WC, waist circumference; BMI, body mass index.

https://doi.org/10.1371/journal.pone.0190862.g002

After we adjusted for confounders, lifestyle intervention was associated with an adjusted odds ratio (OR) 1.33 (95% confidence interval [CI]: 1.31–1.36, p<0.001) for 5% of WC reduction, 1.36 (95% CI 1.33–1.38, p<0.001) for 5% of BMI reduction, and 1.33 (95% CI: 1.31–1.33, p<0.001) for reversal of MetS (Table 3). Confirmatory analyses in a propensity-matched cohort yielded no substantive differences relative to traditional multivariate analyses (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Clinical end point in a whole cohort and a propensity-matched cohort.

https://doi.org/10.1371/journal.pone.0190862.t003

The mean WC changes were −1.34 and −0.44 cm in participants and non-participants, respectively, with a difference of −0.89 cm (95% CI: −0.92 to −0.86). The mean BMI changes were −0.29 and −0.08 kg/m² in participants and non-participants, respectively, with a difference of −0.22 kg/m² (95% CI: −0.22 to −0.21). The intervention program also resulted in significantly greater reductions in both abdominal and overall obesity parameters (S1 Table). Participants, compared to non-participants, had significant reductions in systolic blood pressure (SBP, −1.15 vs −0.72 mm Hg), diastolic blood pressure (DBP, −0.97 vs −0.64 mm Hg), and log TG (−0.11 vs −0.08), respectively (p<0.001 for each). Participants also improved their HDL-cholesterol level more than non-participants (1.48 vs 0.94 mg/dl). Of all parameters for obesity and cardiometabolic risk factors, greater reductions were observed in abdominal and overall obesity profiles, compared to cardiometabolic parameters. The adjusted differences in the traditional adjusted model were attenuated (ΔWC coefficient: −0.76; 95% CI: −0.79 to −0.73; ΔBMI coefficient: −0.19; −0.20 to −0.18), and the coefficients in the propensity-matched cohort were further attenuated (ΔWC: −0.77; −0.82 to −0.72; ΔBMI: −0.20; −0.21 to −0.18; Fig 3). The IV model strengthened the findings of the traditional adjusted model (Fig 3). Similar patterns were found for other cardiometabolic risk factors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Beta-coefficients of each cardiometabolic risk factors using linear regression and instrumental variable methods.

SBP, systolic blood pressure; DBP, diastolic blood pressure; PS, propensity score; IV, instrumental variable.

https://doi.org/10.1371/journal.pone.0190862.g003

Discussion

This study examined the effectiveness of a nationwide preventive program, which is comprised of screening, counselling, and education, using national “real world” data. Nationwide lifestyle intervention is effective in achieving clinically relevant reductions in WC and BMI and reversal of MetS. This intervention was also effective for long-term changes in cardiometabolic risk factors.

MetS is widely used to identify patients with abdominal obesity who have increased risk for CVD and diabetes [14]. While metabolic components likely overlap [15], we previously demonstrated that abdominal obesity is an independent predictor for new onset of individual MetS components in the longitudinal study [16]. In a previous study, abdominal obesity was found to be important in MetS development and preceded the development of other MetS components [17, 18]. Other researchers showed that abdominal obesity is superior to overall obesity for predicting total and cardiovascular mortality rates, suggesting that a reduction in abdominal obesity should reduce the risk of MetS and negative cardiovascular outcomes [19, 20].

A health promotion program focusing on lifestyles can substantially reduce obesity and CVD risk. However, the findings from previous reports have been equivocal. Despite the potentially large benefits of obesity reduction, lifestyle interventions have had small long-term success, especially at the population level [7, 21, 22]. In a previous meta-analysis, lifestyle interventions significantly reduced blood pressure and low-density lipoprotein cholesterol in the general population, but the pooled estimates have dubious validity because of marked unexplained heterogeneity among trials (I² statistic >85%) [23, 24]. In the present study, we have shown long-term improvement in reducing obesity and cardiometabolic risk factors through lifestyle intervention within the framework of nationwide policy. These findings add further real-world evidence that suggests the nationwide preventive program that focuses on abdominal obesity was truly effective.

Although the randomized controlled study (RCT) unquestionably remains a powerful tool to minimize the risk of bias due to confounding factors and to develop scientific evidence, the usefulness of real-world evidence is emphasized through its potential for complementing the knowledge gained from clinical trials [25], where limitations make it difficult to generalize findings to larger nationwide settings [26]. In a previous population-based randomized lifestyle intervention study (the Inter99 study) [27], screening and lifestyle counselling had sustained effects on physical activity and dietary habits after the intervention discontinuation [28] and changes in physical activity level from baseline to 5-year follow-up were associated with relevant changes in weight, WC, DBP and serum lipids [29]. Nevertheless, there was no effect on incidence of diabetes [30] and development of fatal and non-fatal ischemic heart disease in the general population [31]. As the author discussed in the limitations, a 10-year follow-up could be too short to consider CVD. However, it is still early to conclude, based only on the randomized controlled trials [23, 31], if health checks followed by lifestyle counselling in a general population should not be part of a country’s health policy. Traditional RCTs are often conducted with specific populations and in a specialized environment that differ significantly from the realities of a public health policy setting and therefore RCTs might not provide a realistic view of policy impact for the general population [25]. Comprehensive and integrated public health prevention methods ideally attempt to implement multifaceted measures at each level of the health impact pyramid, which consists of five tiers: (1) socioeconomic factors; (2) changing the context to create healthy default decisions in the individual; (3) long-lasting protective interventions; (4) clinical interventions; and (5) counselling and education [32]. The Japanese national health policy is in line with the ideal of clinical care impacting population health. All Japanese people of the targeted age have equal opportunities to receive health checkups and lifestyle interventions under the law (tiers 1 and 5). The annual health checkup enables continuous monitoring (tier 3). Individuals at risk of developing CVD can receive counselling and educational interventions (tier 5), which can connect them to appropriate medical treatment (tier 2). The greatest differences between RCT and a public health policy setting is that the nationwide policy enables almost all citizens to learn the term “metabolic syndrome,” suggesting that it may be able to change the context of health by altering social norms (tiers 4 and 5). Moreover, this action has been conducted at the country level and led by the government, as recommended by the World Health Organization [33]. Therefore, the lifestyle intervention has maximized synergy, and we have achieved long-term success. In our analysis, we provide evidence that by implementing interventions, sustained public health and clinical benefits, as set forth at the beginning of the study, can be achieved. This study can lead to other successful population-based prevention strategies in a national policy context. Thus, the national lifestyle intervention can continue to yield a substantial and sustained reduction in abdominal obesity and cardiometabolic risk factors.

Our study had several limitations. Evaluating the effectiveness of the lifestyle interventions using non-randomized data has been controversial because of potential biases related to treatment selection (attending the lifestyle intervention or not) and the lack of data on potential confounding variables, including socioeconomic status, income, education, health literacy, or health motivation. The previous report suggested that there are two reasons for not attending the lifestyle interventions. Subjects may already have regular office visits, and subjects may not have time to attend the intervention. In the former scenario, non-participants must initiate lifestyle changes at their clinic, resulting in underestimated effectiveness of nationwide intervention. In the latter scenario, non-participants are not always less motivated than participants. Since participation increased more than doubled by shortening the intervention session time in the previous study, some non-participants may be willing to change their lifestyles on their own. Therefore, the latter reason may lead to either underestimation or overestimation of the intervention effect. To mitigate bias, we used three steps to assess intervention effectiveness. First, we controlled for demographic and clinical factors using a traditional multivariate model. Secondly, we used propensity-score matching to further control for selection bias. We applied all possible data believed to potentially affect the decision to attend lifestyle counselling in an effort to maximally reduce selection bias, which carefully balanced observed baseline characteristics [34]. If the distribution of unmeasured factors is more likely to be similar with similar clinical indications and risk when considering therapies, unmeasured factors can closely correlate with measured factors. In this setting, propensity scores may remove bias due to treatment selection differences resulting from unmeasured factors [10]. When strong selection bias exists, however, both multivariate adjustment and propensity-based matching are no more likely to remove bias due to unmeasured confounding variables. Thirdly, we have added IV analyses, which is a method designed to control for unmeasured bias. Our finding that several baseline characteristics modified the comparative effectiveness of lifestyle intervention is also consistent with the results of all analytical steps. Our analyses of nationwide observational data can provide reliable estimates of real-world lifestyle intervention effectiveness [35].

The second limitation includes the natural constraints of national data. Although a consistent format was used to collect data, the reported data were subject to data error and local variability. However, it was unlikely that the groups (i.e., participants and non-participants) would be differentially affected. Third, our subjects were selected based on attending checkups in both the first year and fiscal year 2011, indicating that they were health-conscious individuals. Our follow-up was 3 years, and the effects may attenuate over time. Fourth, we could not evaluate specific program features including individual vs. group sessions or physical activity vs. diet vs. both. Future research needs to identify which features optimize the effectiveness of the programs and which are less critical [36]. Fifth, the smoking status may be under-reported in self-answered questionnaires. Finally, the generalizability of our findings to other countries or ethnic groups with a higher burden of obesity and adverse cardiometabolic diseases is unknown. The aspects of Japanese culture may have contributed to our results, including the factors that already contribute to the very low obesity prevalence in Japan compared to other nations. It can also be difficult to adopt the current estimates worldwide due to a different MetS criteria in Japan. However, the United States Preventive Services Task Force recommends screening all adults for obesity and offering or referring obese adults to intensive behavioral interventions to improve weight status and other risk factors for important health outcomes [2]. In Finland, cardiovascular risk factor levels declined markedly after the implementation of a national CVD prevention strategy during the late 1970s [37]. Recently, the comprehensive healthy lifestyle program, comprising preventive and promotional activities that consider both population and high-risk approaches, has been applied to a middle-income country [38]. These studies introduce possible beneficial effects of nationwide lifestyle interventions in developed and developing countries. Further research is required to evaluate the long-term cost-effectiveness of a nationwide screening and lifestyle intervention. Clinical effectiveness and cost effectiveness underpinned by the best clinical evidence should be discussed in parallel [39].

In conclusion, the national screening and lifestyle intervention yielded a substantial and sustained improvement in abdominal obesity and reversal of MetS.