Introduction

In 2021, diabetes affected approximately 537 million people worldwide1, resulting in about 966 billion US dollars in global health expenditures1. These figures are projected to increase by 20451. Prediabetes is also highly prevalent. For example, ~38% of adults (≥18 years) in the US have prediabetes2. In Japan, among working adults (≥20 years), ~13% meet prediabetes criteria by fasting plasma glucose alone and 20% by hemoglobin A1c (HbA1c) alone3. Prediabetes is considered a high-risk state for progression to type 2 diabetes, as about 10% of individuals with prediabetes develop diabetes each year2. Major phenotypes of prediabetes include impaired fasting glucose and impaired glucose tolerance2. The definition of prediabetes varies slightly among organizations. The American Diabetes Association (ADA) defines prediabetes as fasting plasma glucose of 100–125 mg/dL, 2-h post-load plasma glucose of 140–199 mg/dL, HbA1c of 5.7–6.4%)4; World Health Organization uses a higher fasting plasma glucose threshold (110–125 mg/dL) with the same 2-h post-load plasma glucose range (140–199 mg/dL); and the International Expert Committee criteria defines prediabetes as HbA1c of 6.0–6.4%2. Prediabetes is associated with an increased risks of death, cardiovascular events2, and cancer5, even without progression to diabetes6. Thus, effective low-cost strategies for preventing prediabetes are urgently needed.

Cohort studies have shown that certain lifestyle factors are associated with prediabetes risk. Smoking7 and heavy alcohol drinking8,9 are linked to an increased risk, whereas physical activity10,11, maintaining normal body weight12, and avoiding both short and long sleep duration13 are associated a reduced risk. These findings suggest that adopting healthy lifestyles may play a key role in preventing prediabetes. This is further supported by the fact that lifestyle modification is the first-line therapy for prediabetes2.

However, two important gaps remain in the literature on prediabetes prevention. First, previous studies7,8,9,10,11,12,13 have examined lifestyle factors in isolation, and no studies explored the combined effects of multiple lifestyle behaviors. If adopting a greater number of healthy behaviors is associated with a lower risk of prediabetes, evidence of such a dose–response relationship could provide stronger motivation for individuals to adopt healthier lifestyles. Second, to our knowledge, no research has characterized the trajectories of overall healthy lifestyles among adults with normoglycemia in a real-world setting. Understanding lifestyle trajectory patterns prior to the onset of prediabetes could help health professionals design tailored prevention strategies for specific subgroups. Indeed, increasing attention has been paid to modeling trajectories of health-related conditions in medicine14,15. In this study, we identify trajectory patterns of combined healthy lifestyles among normoglycemic Japanese workers and investigated their associations with incident prediabetes. Our findings suggest that maintaining or improving healthy lifestyles is associated with a lower risk of prediabetes in this working population.

Methods

Study settings

This cohort study used the health checkup from a large-scale company, derived from a sub-study on exercise epidemiology16 within the Japan Epidemiology Collaboration on Occupational Health (J-ECOH) Study6. In Japan, annual health examinations, including blood glucose testing, are mandated for all employees under the Industrial Safety and Health Act. Informed consent from each participant was waived in accordance with the ethical guidelines, where informed consent was not necessarily required for observational studies using existing data. J-ECOH Study and its implementation were announced and explained at the participating company; employees were informed of their right to opt out of research use of their data. The study protocol including the use of the present sub-cohort data, was approved by the Ethics Committee of the National Center for Global Health and Medicine (approval number: NCGM-G-001140).

Study design and participants

An overview of the study design is presented in Fig. 1. Lifestyles were assessed between 2006 and 2009 (over 3 years). The incidence of prediabetes was evaluated from 2009 to 2017. Similar study designs have been employed previously17,18. The study initially included 43,025 employees (36,208 men and 6817 women), aged 30–64 years, who underwent a health checkup fiscal year 2009 (April 2009 and March 2010 baseline). Participant selection is detailed in Supplementary Fig. 1. We excluded 8628 participants who did not have baseline data for diabetes diagnosis (n = 5291) or having diabetes at baseline (n = 3480), as defined by fasting glucose of ≥126 mg/dL, random plasma glucose of ≥200 mg/dL, HbA1c levels of ≥6.5%, ongoing anti-diabetic drug therapy, and self-reported history of diabetes4. We further excluded 19,590 participants with prediabetes at baseline (HbA1c of ≥5.7%, fasting blood glucose of ≥100 mg/dL, n = 18,628) or missing data on fasting plasma glucose (n = 1420). Of the remaining 14,807 normoglycemic individuals, we sequentially excluded 257 participants who reported a current or past history of cancer, myocardial infarction, or stroke, 1704 participants without baseline data on lifestyle factors or covariates, 1690 attending annual health checkups or having data on health-lifestyles at fewer than two time points between 2006 and 2008. Lastly, we excluded due to lack of follow-up data: 383 participants who did not attend any health checkup after the baseline or who did not have data on plasma glucose or HbA1c levels after the baseline. The main analysis included 10,773 participants (8986 men and 1787 women), with a mean age of 43.3 years (range: 30–64 years).

Fig. 1: Study design.
Fig. 1: Study design.
Full size image

Health-related lifestyles were assessed from 2006 to 2009 to determine the trajectories of health-related lifestyles over 3 years among participants without prediabetes and diabetes in 2009. Occurrence of prediabetes was assessed from 2010 to 2017 using annual health checkup data.

Biochemical measurements

Blood samples were collected after an overnight fast of ≥10 h. Blood glucose levels were measured using the glucose electrode technique, and HbA1c levels were determined by high-performance liquid chromatography (HPLC) method.

Lifestyle assessments

Lifestyle information was collected annually via self-administered questionnaire during health checkups. Data included smoking status (never, past, or current), alcohol consumption (in go per day, with 1 go ≈ 23 g of ethanol), sleep duration (<5, 5–6, 6–7, and ≥7 h/day), and exercise. Weekly exercise volume (metabolic equivalent [MET]-hours) was estimated using the self-reported activity type (up to three from 20 items), frequency, and duration18. Body mass index (BMI, kg/m2) was calculated using objectively measured body height (m) and weight (kg).

Development of the healthy lifestyle index

Following prior methodology18, we used an a priori lifestyle index composed of five factors: smoking, alcohol use, sleep, exercise, and body weight. One point was assigned for each low-risk behavior; otherwise, zero. The total score ranged from 0 (unhealthy lifestyles) to 5 (healthy lifestyle). This index was calculated annually from 2006 to 2009 (three–four times). Low-risk criteria were: no smoking (never or former smoker), no heavy alcohol use (<46 g for men; <23 g per day for women), sufficient sleep (≥7 h/day), adequate physical activity (≥7.5 MET-h/week), and healthy body weight (BMI < 25.0 kg/m2).

Outcome assessment

Participants were followed from 2010 to 2017 to capture incident prediabetes. Prediabetes was defined using the ADA criteria4: fasting plasma glucose of 100–125 mg/dL (5.6–6.9 mmol/L), HbA1c of 5.7–6.4% (38–46 mmol/mol). If a participant developed diabetes without meeting the prediabetes criteria beforehand, such a person was considered as having prediabetes at the midpoint between the health checkup date of meeting the diabetes criteria and the last health checkup date before meeting the diabetes criteria. In sensitivity analyses, prediabetes was also defined based on the World Health Organization/ International Expert Committee criteria: fasting plasma glucose of 110–125 mg/dL or HbA1c of 6.0–6.4%4,6.

Other variables at baseline

Participants reported their sociodemographic conditions, work-related status, and personal history of diseases at baseline as previously described18. Work-related status included overtime work hours, shift work, job position, and duration of walking to and from work. Hypertension was defined as systolic blood pressure of ≥140 mmHg, diastolic blood pressure of ≥90 mmHg, or self-reported antihypertensive treatment.

Statistics and reproducibility

We applied group-based trajectory modeling to categorize participants into lifestyle patterns across 2006–2009, guided by Bayesian information criteria and the need to ensure adequate subgroup sizes. This approach has have been detailed previously18. Classification quality was assessed using average posterior probabilities (≥70% indicated good discrimination). Patterns were labeled based on average score and score trajectory.

Descriptive statistics were presented as percentages (categorical variables) and means (continuous variables) by lifestyle patterns. We calculated person-time using the date of the baseline check-up and the date of diagnosis with prediabetes at a subsequent health check-up or the date of the last check-up. We employed Cox regression to obtain hazard ratios and 95% confidence intervals (CIs) of prediabetes onset according to longitudinal lifestyle patterns. We treated the very unhealthy pattern as reference because this pattern showed the highest risk. First, we adjusted for sex and age (years, continuous) at baseline in model 1. In the next model, we further adjusted for hypertension (yes or no), a family history of diabetes (yes or no), job position (high or others), monthly overtime work (<45, 45–59, 60–79, 80–99, and ≥100 h), physical activity at work (sedentary, standing or walking, and fairly physically active), shift work status (yes and no), walking during commute to and from work (<20, 20–39, and ≥40 min). In model 3, we additionally adjusted for baseline HbA1c level (%, continuous) as a potential mediator. Model 4 was adjusted for all covariates in model 2 plus each lifestyle component to examine the mediating role of these lifestyle factors. We considered that 95% CIs of the hazard ratios that do not cross 1.0 as statistically significant. Stata version 18.0 (Stata Corp, College Station, Texas) was employed for data analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Trajectories of healthy lifestyle index and baseline characteristics of participants

Group-based trajectory modeling identified five distinct trajectories of healthy lifestyle index, ranging from a “persistently very unhealthy” trajectory to a “persistently mostly healthy” trajectory (Supplementary Fig. 2). The transitions in individual lifestyle components across these trajectories are detailed in Supplementary Table 1. The majority of the participants were classified into the “persistently moderately healthy” group (n = 4625), followed by the “persistently unhealthy” group (n = 3730). Only 405 participants were in the “improved from unhealthy to moderately health” group. The average posterior probability of assigning participants to each group was sufficiently high, exceeding 90% except for the improved group (83%). As shown in Supplementary Table 2, participants in the healthier lifestyle groups were more likely to be female, hold higher-ranked job positions and work in sedentary jobs and shorter overtime work hours. They were less likely to be shift workers and to report a shorter walking duration during their commute, and they were less likely to have hypertension and family history of diabetes.

Association of healthy lifestyle index trajectories and prediabetes risk

Over a total of 43,732 person-years of follow-up, 6935 participants developed prediabetes (cumulative incidence 64.4%). Only 17 participants progressed directly to diabetes without first meeting the criteria for prediabetes. As shown in Table 1, prediabetes risks were lower in groups with healthier lifestyles. After adjustment for various factors including sociodemographic and work-related factors (model 2), compared with a “persistently very unhealthy” group, the adjusted hazard ratios (95% CIs) were 0.92 (0.85, 1.00) for “persistently unhealthy” group, 0.82 (0.71, 0.95) for “improved from unhealthy to moderately healthy” group, 0.83 (0.76, 0.90) for “persistently moderately healthy” group, and 0.74 (0.67, 0.83) for “persistently mostly healthy” group. Additional adjustment for baseline HbA1c level slightly attenuated the associations. When we adjusted for the baseline lifestyle component (Supplementary Table 3), further adjustment for BMI greatly attenuated; the corresponding HRs were 1.00 (0.92, 1.09), 0.91 (0.79, 1.06), 0.96 (0.88, 1.05), and 0.87 (0.78, 0.98), respectively. Adjustment for other lifestyle factors did not largely change the associations. Similar trajectory patterns and associations were observed in sensitivity analyses using World Health Organization/ International Expert Committee criteria for prediabetes (Supplementary Fig. 3 and Supplementary Table 4).

Table 1 Hazard ratios of incident prediabetes according to trajectories of health-related lifestyles
Full size table

Discussion

Our findings among working adults demonstrated that healthier lifestyle trajectories were associated with a reduced risk of developing prediabetes compared with persistently unhealthy lifestyles. Notably, individuals who transitioned from unhealthy to healthier lifestyles had a lower risk of prediabetes than those with persistently unhealthy patterns. By using longitudinal data on both lifestyle behaviors and prediabetes onset, this study provides new insights into the preventive potential of lifestyle improvements.

The observed risk reductions of prediabetes risk associated with healthier lifestyle trajectories support the previous observational studies demonstrating that individual baseline lifestyle factors are linked to prediabetes risk7,8,9,10,11,12,13. However, these earlier studies7,8,9,10,11,12,13 did not examine longitudinal changes across multiple lifestyle domains. Our findings highlight the importance of tracking multiple lifestyle behaviors over time for effective prediabetes prevention. Despite this, the trajectory modeling revealed that most individuals maintained their lifestyles, underscoring the need to develop and implement scalable, low-cost interventions to facilitate behavior change.

In our study, participants with healthier lifestyles were more likely to refrain from smoking and heavy alcohol use, engage in regular exercise, and maintain a healthy body weight—suggesting that these behaviors may serve as key targets for intervention. At the individual levels, employees should be encouraged to adopt healthier habits, including smoking cessation through pharmacotherapy and/or behavioral counseling19, reduce heavy alcohol consumption via regular screening and, if needed, specialist treatment20, and the adoption of physically active routines such as active commuting21, complemented by dietary modifications22 for weight management. At the workplace levels, it is essential for companies to establish health policies and promote wellness initiatives23. Organizations should also implement smoke-free policies, encourage physical activity, and offer healthier food options to support weight loss among employees23. As overweight and obesity are important risk factors for prediabetes2, as also indicated by our findings, workplace health strategies should incorporate both weight gain prevention among normal-weight individuals and weight reduction among those with overweight or obesity.

We found that individuals who improved their lifestyle behaviors had a lower risk of developing prediabetes compared with those who did not improve their lifestyles in a real-world context. These findings align with a previous cohort study examining lifestyle trajectories and diabetes risk18, as well as randomized trials demonstrating the effectiveness of lifestyle modification in preventing progression from prediabetes to diabetes2. Our study contributes new evidence by showing that an increase in the number of healthy behaviors is associated with reduced risk of progression from normoglycemia to prediabetes in everyday settings. Further research is needed to explore the underlying factors that enable individuals to improve their lifestyles in daily life, which could inform the development of effective behavior change strategies.

Our results suggest that the associations between lifestyle patterns and prediabetes risk were generally robust to differences in the definition of prediabetes. Since the absolute risk of developing prediabetes varies between the criteria of the ADA and World Health Organization/ International Expert Committee, researchers may select the definition that best aligns with their study objectives and the research context.

Several mechanisms may explain the relationship between lifestyle behaviors and glucose metabolism. Unhealthy lifestyles—such as smoking24, heavy alcohol use25, insufficient sleep26, physical inactivity27, and excess body weight28, can exacerbate insulin resistance, leading to impaired glucose uptake. Additionally, these lifestyles may contribute to β-cell dysfunction, resulting in insufficient insulin secretion in response to glucose intake24,28,29,30,31.

This study has some strengths, including the examination of longitudinal adherence to healthy lifestyles and the repeated assessment of glycemic status over the long follow-up period. Identifying distinct lifestyle trajectories may support the development of precision lifestyle medicine approaches32. However, several limitations should be acknowledged. First, diet was not incorporated into the lifestyle index. Second, lifestyle behaviors after the baseline periods were not considered, which may have influenced the glycemic outcomes during follow-up. Third, although prediabetes comprises two phenotypes—impaired fasting glucose and impaired glucose tolerance2—we primarily assessed the former using fasting plasma glucose and HbA1c. Therefore, the applicability of our findings to individuals with impaired glucose tolerance (as defined by oral glucose tolerance test or 2-h post-load plasma glucosee) remains uncertain. Fourth, residual or unmeasured confounding may have influenced the observed associations. Lastly, the study population consisted of Japanese workers with high proportion of men, limiting the generalizability of our findings to populations with different sociodemographic characteristics. While directly comparable longitudinal data are limited, our observed risk of prediabetes appears consistent with the US data indicating that prevalence of prediabetes, as defined by the ADA criteria, is 38% among adults aged 18 years or older2,33, given that about 20% of the individuals with prediabetes reverted to normoglycemia, over 10% of progressed to diabetes, and the remainder remained in the prediabetic state34.

In conclusion, this cohort study suggests that individuals following healthier lifestyle trajectories had reduced risk of developing prediabetes. These findings underscore the importance of promoting lifestyle improvements to prevent prediabetes. Further research is needed to develop effective intervention strategies targeting working adults with unhealthy lifestyle behaviors.