The prevalence of diabetes has been increasing worldwide. The International Diabetes Federation estimates that the number of adults with diabetes was 537 million (one in ten adults) in 2021 and predicts that this will increase to 643 million by 2030 and 783 million by 2045(1). In Japan, it is estimated that 19·7 % and 10·8 % of adult men and women, respectively, have diabetes(2). Type 2 diabetes and its complications decrease quality of life and represent a major healthcare burden worldwide(Reference Khan, Hashim and King3). Therefore, it is important to identify the determinants of type 2 diabetes for prevention. Factors including overweight, physical inactivity and genetics contribute to the development of type 2 diabetes(4). Diet has also received considerable attention as a modifiable factor of type 2 diabetes.
Meat, particularly red meat and processed meat, contains saturated fats, heme Fe, Na, nitrates/nitrite and advanced glycation end products (produced during high-temperature cooking), all of which have been linked to increased risk of type 2 diabetes through insulin resistance, oxidative stress and inflammation(Reference de Oliveira Otto, Alonso and Lee5–Reference Risérus8). Red meat is also a major source of animal protein, essential amino acids, vitamins and minerals. An umbrella review of meta-analyses of prospective studies by Neuenschwander et al. reported that red meat and processed meat intake was associated with an increased risk of type 2 diabetes from evidence rated as high quality(Reference Neuenschwander, Ballon and Weber9). In contrast, another meta-analysis by Zeraatkar et al. concluded that the magnitude of the association between red meat and processed meat intake and type 2 diabetes was very small, and the evidence is of low certainty(Reference Zeraatkar, Guyatt and Alonso-Coello10). The authors of the latter meta-analysis pointed to important limitations of previous studies including inadequate adjustment for known confounders, residual confounding due to observational design and recall bias associated with dietary measurement(Reference Zeraatkar, Guyatt and Alonso-Coello10).
Epidemiological evidence regarding this issue has been mainly derived from studies among Western populations. Given the much lower consumption of red meat in Asia (23·6 kg/capita in Japan and 31·0 kg/capita in China v. 49·0 kg/capita in the USA(11)), the association between meat intake and type 2 diabetes in Asia may also differ. So far, there is a paucity of evidence in Asia with conflicting data, which may be partly ascribed to methodological limitations, including insufficient adjustment of intakes of food other than meat(Reference Liu, Liu and Zhang12,Reference Villegas, Shu and Gao13) and socio-economic status(Reference Kurotani, Nanri and Goto14) and incidence of type 2 diabetes based on self-reported information(Reference Liu, Liu and Zhang12,Reference Kurotani, Nanri and Goto14–Reference Talaei, Wang and Yuan16) . Additionally, in spite of the steady increase in meat intake in Japan (71 g in 1990, 78 g in 2000, 83 g in 2010 and 103 g in 2019(2,17) ), as well as in China and Korea(Reference He, Li and Yang18,Reference Yun, Kim and Oh19) , baseline surveys of the previous studies in Asia were conducted mainly before 2000(Reference Liu, Liu and Zhang12–Reference Kurotani, Nanri and Goto14,Reference Talaei, Wang and Yuan16,Reference Yu, Zheng and Cai20) . Here, we examined the prospective association of total meat, red meat, processed meat and poultry intake with type 2 diabetes in a Japanese working population.
Methods
Study design
The Furukawa Nutrition and Health Study, conducted as part of the Japan Epidemiology Collaboration of Occupation Health Study, is a prospective study among workers of a manufacturing company and its affiliates in Chiba and Kanagawa Prefectures, Japan. The survey was conducted every 3 years: the first to third surveys were conducted in April 2012, April 2015 and April 2018 in Chiba and in May 2013, May 2016 and May 2019 in Kanagawa, respectively. In each survey, participants were asked to fill out two types of survey questionnaire, one specifically designed for diet and the second for overall health-related lifestyle. Additionally, we obtained annual health checkup data, including the results of anthropometric and biochemical measurements and information on the history of diseases, from April 2012 to March 2023. This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human participants were approved by the Ethics Committee of the National Center for Global Health and Medicine, Japan (NCGM-G-001140). Written informed consent was obtained from all participants prior to the survey.
Analytic cohort
The present analysis included 3369 employees who participated in the first, second or third surveys and provided health checkup data. If a participant attended two or more surveys, the first one was regarded as baseline. Of these, we excluded twenty-seven participants who did not complete the questionnaires at baseline. We then excluded 209 participants who reported a history of cancer, CVD, chronic hepatitis, chronic kidney disease, pancreatitis or diabetes at baseline and 157 with type 2 diabetes (fasting blood glucose ≥ 126 mg/dl, casual blood glucose ≥ 200 mg/dl or HbA1c ≥ 6·5 %) at baseline. Some participants had ≥ 2 of these conditions. We further excluded 34 participants with missing data on covariates. Of the remaining 3041 participants, we excluded 310 who did not undergo a health checkup subsequent to baseline and 22 with extremely energy intake (exceeding mean ± 3 s d). Finally, 2709 (2399 men and 310 women aged 18–78 years) were included in the analysis.
Dietary assessment
Dietary habits during the preceding month were assessed using a validated brief self-administered diet history questionnaire(Reference Kobayashi, Honda and Murakami21,Reference Kobayashi, Murakami and Sasaki22) . Dietary intake for fifty-eight food and beverage items, energy and selected nutrients were estimated using an computer algorithm for the brief self-administered diet history questionnaire, with reference to the Standard Tables of Food Composition in Japan(23). Meat intake was estimated using four items of pork and beef; chicken; ham, sausages and bacon and liver. For each item, participants were asked to denote their consumption frequency by choosing one of seven options ranging from none to ≥ 2 times/d. According to a validation study of the brief self-administered diet history questionnaire using 16-d weighed dietary records as the gold standard, Spearman’s correlation coefficients for energy-adjusted intake of total meat (by the density method) in ninety-two men and ninety-two women aged 31–76 years were 0·44 and 0·63, respectively(Reference Kobayashi, Murakami and Sasaki22).
Ascertainment of type 2 diabetes
Type 2 diabetes was confirmed by annual health checkups from baseline to March 2023. Type 2 diabetes was defined by any of a fasting blood glucose ≥ 126 mg/dl, casual blood glucose ≥ 200 mg/dl, HbA1c ≥ 6·5 %, self-report of diabetes or current use of anti-diabetic drugs, with reference to the report of the Committee of the Japan Diabetes Society on the Diagnostic Criteria of diabetes mellitus(Reference Seino, Nanjo and Tajima24). Although we did not obtain information on the type of diabetes, we could reasonable assume that most cases were type 2 diabetes considering the age of the study participants (median 41, range 18–78 years old).
Other variables
The questionnaire was also used to obtain information on smoking, alcohol consumption, physical activity during work and housework or while commuting to work and leisure-time physical activity. Physical activity during work and housework or when commuting was inquired about in terms of time spent doing heavy physical work and standing. Physical activities in leisure-time were expressed as the sum of the metabolic equivalent multiplied by the duration of time (in hours) across each level of physical activity. Information on history of hypertension and dyslipidaemia, body height,\ and body weight were obtained from health checkup data. BMI was calculated as body weight in kilograms divided by the square of body height in metres (kg/m2).
Statistical analysis
We calculated person-years of follow-up for each participant, starting from the date of the health checkup at which they first participated in the survey until the date of the health checkup when type 2 diabetes was confirmed, or the date of their last health checkup.
Participants were divided into tertiles by energy-adjusted intake (by the density method) of total meat, red meat, processed meat and poultry at baseline. Baseline characteristics according to tertile of total meat intake were expressed as mean and s d for continuous variables and percentage for categorical variables. Trend association between confounding factors and total meat intake was tested using linear regression analysis for continuous variables by treating the median in each tertile of total meat intake as a continuous variable and using the Mantel–Haenszel chi-squared test for categorical variables.
Cox proportional hazards analysis was conducted to estimate hazard ratios of type 2 diabetes for tertiles of total meat, red meat, processed meat and poultry intake, with the lowest category as reference. The first model was adjusted for age (years) and sex, and the second model was further adjusted for smoking status (never-smoker, quitter or current smoker consuming < 20 cigarettes/d or ≥ 20 cigarettes/d), alcohol consumption (non-drinker or drinker consuming < 23 g, 23 to < 46 g, or ≥ 46 g of ethanol/d), leisure-time physical activity (metabolic equivalent-hour/week, quartile), physical activity during work and housework or while commuting to work (heavy physical work ≥ 30 min/d or heavy physical work < 30 min/d and standing < 30 min, 30–59 min, or ≥ 1 h/d), hypertension (yes or no), dyslipidaemia (yes or no) and total energy intake (kcal). The third model was further adjusted for intake of rice (g/1000 kcal), vegetables (g/1000 kcal) and fish and shellfish (g/1000 kcal). Since obesity might have been an intermediate pathway in the development of type 2 diabetes, BMI (kg/m2) was added to the third model (model 4). Trend associations were assessed by treating the median in each tertile of meat intake as a continuous variable. We repeated the analysis by creating tertile based on crude intakes of meat instead of energy-adjusted ones. Moreover, to examine the type 2 diabetes risk in association with much higher meat intake, we divided participants with the highest tertile of meat intake into two groups based on the median intake and repeated the analysis.
We also analysed the association between meat intake and type 2 diabetes by age (< 41 [median] years old or ≥ 41 years old) and BMI (< 25 kg/m2 or ≥ 25 kg/m2). We tested the interaction by creating an interaction term of two exposure variables and adding to the model. Since there were small number of women (11 %), we also examined it in men only.
We conducted following sensitivity analyses. First, we additionally adjusted for family history of diabetes among participants who reported family history of diabetes at either first or third survey (n 2528). Second, we additionally adjusted for LDL-cholesterol, HDL-cholesterol and TAG among participants with available data (n 2074). Third, we repeated the analysis while censoring the participant who skipped a health checkup for three or more consecutive years at the last health checkup before such skip.
We checked the proportional assumptions using Schoenfeld residuals and found the proportional hazard assumption was satisfied for all variables in the model 4 except sex. We confirmed that the results were materially unchanged after excluding women (only 11 % of the participants). Two-side P values < 0·05 were considered to indicate statistical significance. All analyses were conducted using Statistical Analysis System (SAS) software version 9.4 (SAS Institute).
Results
The baseline characteristics of study participants according to total meat intake are shown in Table 1. Compared to those with lower intake, participants with higher intake of total meat were younger; less likely to be men, and alcohol drinkers; and less likely to report heavy physical work ≥ 30 min/d and a history of hypertension. They also had higher intake of vegetables and fish and shellfish, but lower intake of rice.
Baseline characteristics of participants according to tertile (T) of total meat intake (Numbers and percentages; mean values and standard deviations)
* Based on the Mantel–Haenszel chi-squared test for categorical variables and linear regression analysis for continuous variables, assigning median to tertile of total meat intake.
† Leisure-time physical activity of ≥ 10·5 METs-hour/week (highest quartile).
During 16 119 person-years (mean 6 years) of follow-up, 135 (5·0 %) participants were newly identified as having type 2 diabetes. The hazard ratios of type 2 diabetes according to tertile of meat intake are shown in Table 2. Total meat intake was not associated with type 2 diabetes in any model. After adjustment for covariates, including dietary factors and BMI (model 4), the hazard ratio of type 2 diabetes for the highest v. lowest tertile was 1·01 (95 % CI 0·63, 1·62) (P for trend = 0·93). In analysis by type of meat, intake of neither red meat, processed meat nor poultry was associated with type 2 diabetes. When we repeated the analysis using crude intake of meat, we again observed no association between meat intake and type 2 diabetes (online Supplementary Table 1). Moreover, when we divided participants with the highest intake group into two group (lower half and higher half), higher level of meat intake was not associated with type 2 diabetes. The multivariable-adjusted hazard ratios of type 2 diabetes for the higher-half intake group were 1·09 (95 % CI 0·59, 2·01) for total meat, 0·95 (95 % CI 0·54, 1·66) for red meat, 1·23 (95 % CI 0·76, 2·00) for processed meat and 1·45 (95 % CI 0·84, 2·52) for poultry compared with the lowest tertile of meat intake.
Hazard ratio (95 % confidence interval) of type 2 diabetes according to tertile (T) of meat intake (Hazard ratios and 95 % confidence intervals)
* Based on Cox proportional hazard regression with assignment of the median in each tertile of meat intake.
† Adjusted for age (years) and sex.
‡ Additionally adjusted for smoking status (never-smoker, quitter or current smoker consuming < 20 cigarettes/d, or ≥ 20 cigarettes/d), alcohol consumption (non-drinker or drinker consuming < 23 g, 23 to < 46 g or ≥ 46 g of ethanol/d), leisure-time physical activity (metabolic equivalent-hour/week, quartile), physical activity during work and housework or while commuting to work (heavy physical work ≥ 30 min/d or heavy physical work < 30 min/d and standing < 30 min, 30–59 min or ≥ 1 h/d), hypertension (yes or no), dyslipidaemia (yes or no) and total energy intake (kcal).
§ Additionally adjusted for intakes of rice (g/1000 kcal), vegetables (g/1000 kcal) and fish and shellfish (g/1000 kcal).
|| Additionally adjusted for BMI (kg/m2).
In stratified analyses by age and BMI, total meat, red meat, processed meat and poultry intake were not associated with type 2 diabetes in any group (Table 3). Similar results were also observed among men only. When we repeated the analysis by additionally adjusting for family history of diabetes and LDL-cholesterol, HDL-cholesterol and TAG and censoring in the previous year those participants who did not have a health checkup for three or more consecutive years, the results were not materially changed (data not shown).
Hazard ratio* (95 % confidence interval) of type 2 diabetes according to tertile (T) of meat intake by age and BMI and men only (Hazard ratios and 95 % confidence intervals)
* Adjusted for age (years), sex, smoking status (never-smoker, quitter or current smoker consuming < 20 cigarettes/d, or ≥ 20 cigarettes/d), alcohol consumption (non-drinker or drinker consuming < 23 g, 23 to < 46 g, or ≥ 46 g of ethanol/d), leisure-time physical activity (metabolic equivalent-hour/week, quartile), physical activity during work and housework or while commuting to work (heavy physical work ≥ 30 min/d or heavy physical work < 30 min/d and standing < 30 min, 30–59 min, or ≥ 1 h/d), hypertension (yes or no), dyslipidaemia (yes or no), total energy intake (kcal), intakes of rice (g/1000 kcal), vegetables (g/1000 kcal) and fish and shellfish (g/1000 kcal), and BMI (kg/m2). In stratified by sex, sex was excluded from adjustment factors.
† Based on Cox proportional hazard regression with assignment of the median in each tertile of meat intake.
Discussion
In this Japanese working population, total meat, red meat, processed meat and poultry intake was not associated with risk of type 2 diabetes. Meat intake was not associated with type 2 diabetes risk in any subgroup stratified by sex, age or BMI. To our knowledge, this study is one of the few to have examined the association of meat intake and type 2 diabetes among an Asian population.
Our finding of no association between red meat and processed meat intake and type 2 diabetes risk is inconsistent with the findings of meta-analyses, showing an increased risk of type 2 diabetes associated high intake of red meat and processed meat(Reference Neuenschwander, Ballon and Weber9,Reference Zeraatkar, Guyatt and Alonso-Coello10,Reference Schwingshackl, Hoffmann and Lampousi25) . A meta-analysis of five Asian studies, where meat consumption is lower, also reported an increased risk among those who consumed higher amount of processed meat and red meat(Reference Yu, Zhang and Xie26). In spite of the consistent associations, the authors of one meta-analysis revealed that the meat-diabetes association was of low certainty to the potential for inadequate adjustment for known confounders, residual confounding and recall bias in dietary measurement(Reference Zeraatkar, Guyatt and Alonso-Coello10). For example, meat intake correlates with income level(Reference Villegas, Shu and Gao13,Reference Son, Lee and Park15,Reference Yu, Zheng and Cai20,Reference Du, Guo and Bennett27) . Participants with high incomes (high meat intake) tend to receive health checkups and visit hospitals and are, therefore, more likely to be diagnosed with type 2 diabetes than those with low incomes (low meat intake), leading to an ascertainment bias in studies that defined the incidence of type 2 diabetes based solely on self-report. In the present study among workers with relatively homogeneous social backgrounds, which defined type 2 diabetes according to both laboratory (blood glucose and HbA1c) and self-reported information and adjusted for a wide range of dietary factors, including rice, vegetables and fish and shellfish intake, we did not find any evidence supporting association of red meat and processed meat with type 2 diabetes.
The no association between red meat and processed meat intake and type 2 diabetes we observed might be partly explained by their lower consumption. In the present study, the median intake in the highest tertile of red meat intake was 26·5 g/1000 kcal. The only Japanese study observed a significantly increase in risk of type 2 diabetes in men with the highest quartile of red meat intake (median intake, 82·7 g/d) compared with those with the lowest quartile, but not in men with the second (31·4 g/d) and third (49·0 g/d) quartile of red meat intake(Reference Kurotani, Nanri and Goto14). Another Asian study also reported an increased risk of type 2 diabetes in the highest quartile of red meat intake, though the intake in that category was not high (median intake, 43·5 g/d)(Reference Talaei, Wang and Yuan16). The Japanese study (in women) and a Chinese study have observed no association, despite high intake of red meat in the highest category (median intake for the highest quartile in the Japanese study, 72·2 g/d(Reference Kurotani, Nanri and Goto14); cut-off point for the highest quintile in the Chinese study, 67·6 g/d(Reference Villegas, Shu and Gao13)). Given the previous findings, the amount of red meat intake might not explain the association with type 2 diabetes. Alternatively, the type of red meat consumed, cooking methods, overall dietary patterns, and lifestyle might contribute to the association between red meat intake and type 2 diabetes. In Japan, in the past several decades, the type of red meat consumed has changed; beef consumption has not almost changed (18·8 g/d in 1990 and 15·2 g/d in 2019), whereas pork consumption has increased (24·8 g/d in 1990 and 40·5 g/d in 2019)(2,28) . In the USA study(Reference Schulze, Manson and Willett29), beef intake was associated with increased risk of type 2 diabetes, whereas pork intake was not associated. In the present study, although we could not examine separately red meat into pork and beef, we might observe no association due to the high proportion of pork in red meat.
Our finding that poultry intake was not associated with type 2 diabetes is consistent with the results of a meta-analysis of 11 prospective studies(Reference Yang, Li and Wang30). Poultry contains a lower amount of fat and heme iron than red meat. A Chinese study reported that poultry intake was associated with decreased risk of type 2 diabetes(Reference Villegas, Shu and Gao13). The authors noted that poultry consumption has been indirectly linked to lower risk of type 2 diabetes; specifically, high consumption of poultry was found to be part of a ‘prudent dietary pattern’ that has been associated with a lower risk of type 2 diabetes. Among Japanese studies, one reported that poultry intake was associated not with a ‘prudent dietary pattern’ but with a ‘Westernized dietary pattern’(Reference Nanri, Shimazu and Takachi31); a second reported that it was associated not with a ‘healthy Japanese dietary pattern’ but with an ‘animal food pattern’(Reference Nanri, Kimura and Matsushita32) and a third reported that it was associated not with a ‘healthy pattern’ but with a ‘high-fat pattern’(Reference Nanri, Yoshida and Yamaji33). Given these findings, poultry intake might not be associated with increased or decreased risk of type 2 diabetes among Japanese.
Strengths of the present study include its prospective design, high study participation rate, use of a validated questionnaire for diet and adjustment for known and suspected risk factors of type 2 diabetes. Our study also has some limitations. First, since meat intake was evaluated using a self-administered questionnaire, misclassification due to measurement error is possible. Second, we used dietary intakes at only one time point and thus could not reflect dietary changes over follow-up. Third, since we did not obtain information on the type of diabetes, we could not rule out the possibility of type 1 diabetes. Fourth, the event rate of type 2 diabetes ranged from 2·8 % to 11·5 % in the previous Asian studies(Reference Zeraatkar, Guyatt and Alonso-Coello10–Reference Liu, Liu and Zhang12,17–Reference Yun, Kim and Oh19) , and it was relatively low in our study (5·0 %). Fifth, although we adjusted for important risk factors for type 2 diabetes, we cannot rule out the possibility of bias due to residual confounding or unmeasured confounding such as inflammatory markers. Finally, because the study participants were workers of a selected company, the present findings may not be applicable to a general population.
In conclusion, we observed no prospective association of total meat, red meat, processed meat or poultry intake with type 2 diabetes. Our findings did not support the hypothesis that higher meat intake is associated with increased risk of type 2 diabetes among Japanese. Considering that Japanese consume less meat than Westerner, and evidence that total meat, red meat and processed meat intake is associated with an increased risk of type 2 diabetes Western studies, we should be cautious increasing meat intake.
Acknowledgements
The authors thank Hiroko Tsuruoka, Rie Ito and Akiko Makabe (Furukawa Electric Corporation) and Yuriko Yagi (National Center for Global Health and Medicine) for their help with data collection. This study was supported by the Industrial Health Foundation (T.M.); JSPS KAKENHI (T.M., grant number JP25293146), (A.N., grant numbers JP25702006, JP21K02133); and the National Center for Global Health and Medicine (T.M., grant number 22A1008).
A.N. and T.M. designed the study; A.N., T.K., I.K., M.K. and T.M. conducted the survey; A.N. analysed data; T.M. provided statistical expertise; A.N. and S.I. wrote the paper. A.N. and T.M. had primary responsibility for final content. All authors read and approved the final manuscript.
There are no conflicts of interest.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114524001922
