Sleep, a circadian pacemaker, is a principal modulator of various metabolic functions, including hormone release and glucose regulation.^1,2 Recent changes in lifestyle factors, such as increased work hours, night activities, or shift work, can reshape sleep characteristics.³ Accumulating evidence indicates that abnormal sleep duration and altered meal timing are independent risk factors for metabolic disorders.^4-6 Of note, sleep duration among the general public has declined concomitantly with the rise of obesity,^7-9 which is widely known to be an indicator of metabolic disorders and has been linked to both morbidity and mortality from diabetes, several cancers, and cardiovascular disease.^10-12

Numerous epidemiological studies, including meta-analyses and cohort studies, have reported an association between chronic circadian rhythm disruption and the risk of metabolic disorders, including obesity, diabetes, metabolic syndrome, and dyslipidemia.^2,5,13-17 However, the currently available evidence contains inconsistent results about the association between sleep duration and metabolic biomarkers, particularly lipid profiles, energy storage, and the regulation of endocrine function.^18,19 A recent meta-analysis of twelve prospective studies found a significant association between the risk of obesity and short sleep duration, but no association was found with long sleep duration.²⁰ However, no supportive evidence was found in another meta-analysis of eight articles, partly due to heterogeneity across studies and publication bias.¹⁸ Various studies have reported positive, inverse, U-shaped, and no association between habitual sleep duration and insulin resistance, with the results varying across populations and by biomarkers assessed.^21-25 A Korean study of 13,609 adults from the 2010 to 2012 Korean National Health and Nutrition Examination Survey found a statistically significant association between sleep duration of ≥9 hours and low levels of high-density lipoprotein cholesterol (HDL-C), compared with 7 hours of sleep, with an odds ratio (OR) of 1.30.²³ Hypertriglyceridemia was associated with inadequate sleep duration in a Japanese cross-sectional study.²⁶

Meanwhile, a growing migrant population is reshaping Korean society by increasing ethnic variety, which could prospectively affect the burden of disease. Among the 16,666 international marriages reported in 2022 in Korea, marriages between a foreign woman and a Korean man accounted for 72% (12,007).²⁷ The number of Filipino immigrant women married to Korean men has increased continuously and now ranks 4th in population size, following those married to men in China, Vietnam, and Japan.²⁸ Given that the physical and social well-being of immigrants is vital for both the immigrants and the host society, understanding and exploring the determinants of immigrant health status is growing in importance. A study of 287 immigrant women in Korea reported that Filipino women have the highest prevalence of obesity, compared with immigrants from China and Vietnam.²⁹ An analysis of data from the U.S. National Health Interview Survey reported that Filipinos were more likely than other Asian ethnic groups to be obese and hypertensive.³⁰ Previous studies about sleep duration examined various factors that mediate sleep behavior such as socio-demographic variables, employment, and acculturation,^13,31,32 which suggests the need for a study that considers the heterogeneity of the immigrant population. In an analysis of Hispanic migrants in the United States, abnormal sleep duration was associated with higher odds of depression and neighborhood environment.³¹ To our knowledge, no researchers have yet taken an epidemiological approach to the association between usual sleep hours and metabolic disorders among an immigrant population in Korea. In this cross-sectional study, we used data from the Filipino Women’s Diet and Health Study (FiLWHEL) to investigate the association between sleep duration and metabolic disorders, along with possible mediating factors.

The Filipino Women’s Diet and Health Study

FiLWHEL is a prospective cohort study of married Filipino immigrant women in the Republic of Korea. From March 2014 to April 2016, 504 baseline participants were recruited from throughout rural and urban areas of Korea: Seoul, Incheon, Daejeon, and rural areas of Gyeonggi and Chungcheong provinces. The participants had to meet two prerequisite conditions: aged 19 years or older and currently or previously married to a Korean man. Further details of the FiLWHEL study have been described elsewhere.³³ The three major components used for data collection were questionnaires, anthropometric measurements, and bio-specimen collection. The FiLWHEL study investigated comprehensive changes in health, social, and dietary factors among immigrant Filipino women from a long-term perspective as they adapted to Korean society. Most of the questionnaires were answered through on-site or telephone interviews. Questionnaires encompassed demographic, socioeconomic, and acculturation-related inquiries. To ensure data quality, the study was assisted by Filipino volunteers who provided clear communication in the Filipino language. Responses were immediately reviewed and verified on-site before the participants were discharged. Anthropometric measurements such as height, weight, and waist hip circumference were collected directly on-site in accordance with the study protocol. We obtained written informed consent from all study participants. This study was approved by the Institutional Review Board of Sookmyung Women’s University (SMWU-1311-BR-012).

Data collection

Of the 504 baseline participants in the FiLWHEL study, we analyzed data for 410 participants after excluding respondents who did not provide complete data for sleep hours (n=15), anthropometric measurements (n=8), or blood samples (n=11); those who were currently pregnant (n=19) or breastfeeding (n=56); and those who reported an energy intake outside the plausible range (>±3 standard deviations) (n=13). Participants were further excluded during the analysis of each outcome if corresponding data were not available. As part of the questionnaire, sleep duration was assessed using the following question: “How many hours do you usually sleep each night/day (if you do night shift work)?” The respondents had four options: <5, 5–6, 7–8, and >8 hours.

Participants’ height was measured directly using a non-stretchable tape, and weight was measured directly using a digital scale or bioelectric impedance analysis machine (InBody 620; Biospace Co. Ltd). Body mass index (BMI, kg/m²) was calculated by dividing weight in kilograms by height in meters squared. A sphygmomanometer was used to measure blood pressure, and the average of two consecutive readings was used. For blood sampling, the participants fasted for at least 8 hours, and blood was drawn by a professional phlebotomist and then immediately centrifuged and refrigerated. The serum levels of triglycerides, total cholesterol, fasting blood glucose, and HDL-C were measured using a Cobas 8000 C702-I (Roche Diagnostics). Glycosylated hemoglobin (HbA1c) levels were measured using high performance liquid chromatography principles on a Tosoh G8 (Tosoh Bioscience). Insulin levels were measured using a Cobas 8000 E602 (Roche Diagnostics) with electrochemiluminescence immunoassay methods. The intra-assay coefficient of variation for each biomarker was 1.21% to 2.79% for fasting blood glucose, 1.49% to 2.99% for total cholesterol, 1.48% to 2.33% for blood triglycerides, 0.98% to 2.11% for HDL-C, and 1.97% to 2.13% for insulin. Low-density lipoprotein cholesterol (LDL-C) was indirectly calculated via the Friedwald formula as follows: LDL-C=total cholesterol–HDL-C–(triglycerides/5).³⁴

The original homeostatic model assessment model from Matthews et al.³⁵ was applied to calculate the homeostasis model assessment of β-cell function (HOMA-β) and insulin resistance (HOMA-IR) estimates from fasting blood glucose and insulin measurements. The HOMA-β indices were calculated as follows: 20×fasting insulin (μIU/mL)/fasting glucose (mmol/mL)–3.5. The HOMA-IR indices were calculated as follows: fasting insulin (μIU/mL)×fasting glucose (mmol/mL)/22.5. Logarithmic transformation was applied to all measurements to follow the normal distribution.

Age, education level, smoking status, coffee intake, and vigorous physical activity data were obtained from structured questionnaires. Details on alcohol intake during the past year were noted. Energy intake was calculated from a single-day 24-hour recall, with portion sizes measured using food miniatures, photographs, household measures, and standard units and portions. The dietary data were computed via CAN-pro 4.0 (Computer Aided Analysis Program 4.0 for professionals; Korean Society of Nutrition). Information missing due to cultural differences was added from international resources: food composition tables from the Food and Nutrition Research Institute of the Philippines,³⁶ Korean Rural Development Administration,³⁷ and U.S. Department of Agriculture.³⁸ Overall health-related quality of life was scored using the World Health Organization Quality of Life-BREF. Alcohol consumption was derived as ethanol intake from the collective intake of soju, beer, liquor, wine, rice wine, and refined rice wine during the previous year. The ethanol intake in grams per day was calculated using the percentage of alcohol in each liquor. Metabolic syndrome criteria from the National Cholesterol Education Program Adult Treatment Panel III were used for the cut-off of each biomarker: waist circumferences ≥88 cm, triglycerides ≥150 mg/dL, HDL-C ≤50 mg/dL, systolic blood pressure (SBP) ≥130 mmHg and diastolic blood pressure (DBP) ≥85 mmHg (hypertension), and fasting blood glucose ≥100 mg/dL.³⁹ A diagnosis of metabolic syndrome was defined as the co-occurrence of three or more of those criteria.

Statistical analysis

The study population was divided into four categories corresponding to their sleep duration (<5, 5–6, 7–8, and >8 hours). Characteristics were compared among the four categories of sleep duration using means with standard deviations or frequencies. The nutrient residual model was applied to calculate energy-adjusted carbohydrate intake.⁴⁰ Both general linear models (GLMs) and logistic regression models were used to evaluate the association between sleep duration and serum levels of triglycerides, HDL-C, LDL-C, total cholesterol, fasting glucose, insulin, HOMA-β, HOMA-IR, BMI, waist circumference, SBP, and DBP. Least squares (LS) means with 95% confidence intervals (CIs) were calculated for each category of sleep duration using GLM procedures. The multivariable model was adjusted for education level (associate/vocational or less, college graduate or more), vigorous physical activity (yes, no), smoking (never, ever), energy intake (kcal/day, continuous), carbohydrate intake (g/day, continuous), health-related quality of life (continuous), and coffee intake (<0.5, 0.5–<2, ≥2 cups/day). We additionally adjusted for BMI (≤23, 23–25, >25 kg/m²), a possible intermediate factor, in the further multivariable model for the lipid and insulin resistance analyses. We conducted pairwise post hoc testing for each category of sleep duration using the contrast statement and Tukey multiple comparison test. A logistic regression model was applied to generate ORs and 95% CIs and included the same covariates used for the GLM in each model. Tests for trends were performed using an ordinal variable for sleep duration.

The other covariates under consideration were menopausal status (yes, no), nap time (minutes, continuous), and length of residence in Korea (years, continuous), but including them in the model did not significantly change the associations; therefore, those covariates were not included in the final model. An interaction analysis was performed using the Wald-test of the cross-product term. We further examined whether the association between sleep duration and circulating triglyceride levels varied by age (≤median or >median), BMI (<25, ≥25 kg/m²), energy intake (≤median or >median), current alcohol drinking status (yes, no), and smoking status (never, ever). All analyses were conducted in SAS version 9.4 (SAS Institute Inc.). All statistical tests were two-sided, and P-values less than 0.05 were considered statistically significant.

From the 504 baseline participants of the FiLWHEL study, we analyzed 410 participants for the blood pressure analysis, 404 participants for the obesity analysis, and 401 participants for the lipid profile and insulin resistance analyses. The baseline characteristics of the 410 study participants are presented in Table 1 according to their habitual sleep duration. Compared with individuals with the shortest sleep duration, those with the longest sleep duration were slightly younger, had higher but moderate alcohol intake, greater levels of LDL-C, and had stayed for a shorter duration in Korea. Those with >8 hours of sleep duration were less likely to have ever smoked or drink coffee than those with <5 hours of sleep duration.

We found that increasing sleep duration was associated with circulating levels of blood triglycerides, total cholesterol, and LDL-C (Table 2). The LS means for <5, 5–6, 7–8, and >8 hours of sleep were as follows: for triglycerides, 81.74 mg/dL (95% CI, 71.43 to 93.54), 85.15 mg/dL (95% CI, 76.65 to 94.59), 86.33 mg/dL (95% CI, 77.84 to 95.75), and 105.22 mg/dL (95% CI, 88.07 to 125.71), respectively (P trend=0.049); for total cholesterol, 174.52 mg/dL (95% CI, 165.02 to 184.57), 180.50 mg/dL (95% CI, 172.79 to 188.55), 182.51 mg/dL (95% CI, 174.83 to 190.53), and 190.16 mg/dL (95% CI, 176.61 to 204.74), respectively (P trend=0.042); and for LDL-C 97.34 mg/dL (95% CI, 88.80 to 106.71), 100.69 mg/dL (95% CI, 93.73 to 108.18), 104.47 mg/dL (95% CI, 97.35 to 112.10), and 109.43 mg/dL (95% CI, 96.94 to 123.54), respectively (P trend=0.047). Tukey’s post hoc analysis was significant at the P<0.05 level for the multiple comparison between mean triglyceride levels with >8 hours of sleep and each of the other categories of sleep duration. The multiple comparison between the mean total cholesterol with >8 hours of sleep and <5 hours of sleep was also significant.

When we additionally adjusted for BMI, the association remained significant for triglyceride and total cholesterol levels. The LS means for <5, 5–6, 7–8, and >8 hours of sleep were as follows: for triglycerides, 81.75 mg/dL (95% CI, 71.68 to 93.25), 83.59 mg/dL (95% CI, 75.48 to 92.56), 86.74 mg/dL (95% CI, 78.49 to 95.86), and 102.26 mg/dL (95% CI, 86.12 to 121.42), respectively (P trend=0.040) and for total cholesterol 174.67 mg/dL (95% CI, 165.10 to 184.78), 179.69 mg/dL (95% CI, 172.02 to 187.71), 182.51 mg/dL (95% CI, 174.87 to 190.48), and 188.59 mg/dL (95% CI, 175.23 to 202.98), respectively (P trend=0.049). We did not observe any significant associations between sleep duration and BMI, waist circumference, HDL-C, fasting glucose, HbA1c, insulin, HOMA-IR, HOMA-β, SBP, or DBP (Tables 2 and 3). A similar trend was observed after additionally adjusting for blood pressure, alcohol intake, and fasting glucose (Supplementary Table 1) or excluding participants who consumed more than 20 g of alcohol per day (Supplementary Table 2). The association was also maintained after excluding participants who were taking medications or previously diagnosed with hyperlipidemia or diabetes (Supplementary Table 3).

Participants with >8 hours sleep duration had 4.47 times higher odds (95% CI, 1.52 to 13.21; P trend=0.033) of high triglycerides than participants who slept for 7–8 hours per night (Table 4). We further conducted a subgroup analysis for circulating levels of triglycerides and total cholesterol (Table 5). The positive association with triglyceride levels was more pronounced in alcohol drinkers than in non-drinkers (P interaction=0.048). No significant interactions were found for total cholesterol.

In this study, long sleep duration was associated with increased serum triglycerides, total cholesterol, and LDL-C levels among Filipino women in Korea. These associations were independent of age, BMI, smoking status, alcohol drinking, health-related quality of life, education, and coffee intake. However, we did not observe any significant associations between sleep duration and obesity, insulin resistance markers, or blood pressure.

Several longitudinal and cross-sectional studies reported that sleep curtailment or surplus was associated with an increased risk of obesity,^41,42 insulin resistance,^43,44 or metabolic syndrome.^5,45,46 Our study finding is consistent with a few epidemiologic studies reporting that increased sleep duration is associated with an elevated risk of metabolic disorder biomarkers. The U.S. Multiethnic Cohort Study²¹ and Chinese Guangzhou Biobank Cohort Study⁴⁷ observed that sleep duration was positively associated with circulating levels of triglycerides. Meanwhile, a U-shaped association between sleep duration and triglyceride levels was reported in Japanese^24,26 and Chinese studies.⁴⁸ Several studies found no association between sleep duration and triglyceride levels.^22,25,49 As for circulating levels of cholesterol, the Coronary Artery Risk Development in Young Adults Sleep study²³ and the Rotterdam study⁵⁰ reported that long sleep duration was associated with increased levels of total cholesterol, whereas other studies found an inverse^25,51 or null association.²² A few studies reported mixed results regarding the relationship between HDL-C or LDL-C and sleep duration. Low^21,22,25,26 or similar HDL-C levels⁴¹ were found with long sleep duration, compared with short or normal sleep duration, and low^26,42 or similar^22,25,49 LDL-C levels were also found with long sleep duration, compared with short or normal sleep duration, in cross-sectional and prospective studies.

Our study showed discrepant findings with regard to obesity and insulin resistance, although the reasons are unclear. For blood pressure, a recent meta-analysis showed that short sleep duration was associated with hypertension.^45,46,52 However, in our study, participants with <5 hours of sleep duration had lower blood pressures than those in the upper categories of sleep duration. Although we did not find a significant association with obesity, our results might suggest a potential link to metabolic syndrome.

Our results further suggest that the association with triglyceride levels could vary by alcohol drinking status. Alcohol drinking elevates triglyceride levels, so ethanol intake might modify the association between sleep duration and triglyceride levels. However, the alcohol consumption of the Filipino participants was modest in quantity, and the association was maintained after adjusting for alcohol intake. In the stratified analysis, a more pronounced association was found for current drinkers than non-drinkers. Given that more than 90% of the drinking population consumed less than 20 g of alcohol per day, moderate alcohol intake among the study participants was correlated with lower triglyceride levels in our results. That finding aligns with previous epidemiologic studies, including the Copenhagen City Heart study, which reported a J-shaped association between alcohol intake and triglycerides in women.⁵³ A recent Mendelian randomization study also reported an association between moderate alcohol consumption and lower triglyceridemia.⁵⁴ The inconsistent results from current and previous research suggests the need for further investigation of this potential interaction with moderate and heavy alcohol drinking in multiple populations.

Although the underlying mechanism between short sleep duration and the risk of metabolic disorders has been suggested to be linked to the appetite regulation of leptin and ghrelin, appetite suppressor and stimulant hormones, respectively,^55-57 the etiology behind the association between long sleep duration and an abnormal metabolic pathway is relatively unknown. Excessive sleep duration could decrease myokine production, impair the phase coherence between circadian regulation and behavioral rhythms such as meal timing, and increase sedentary behavior and physical inactivity. Such behavioral and hormonal changes could result in decreased glucose effectiveness, systemic insulin sensitivity, and synthesis of the satiety-enhancing factor glucagon-like peptide-1 and result in a higher risk of obesity and type 2 diabetes mellitus.⁵⁸ Of note, recent studies have suggested a bidirectional relationship among metabolic disorders with a positive feedback cycle that enhances abnormal sleep duration.^3,58

The strength of our findings is that this is the first study, to our knowledge, to examine the relationship between sleep duration and metabolic abnormalities among Filipino women in Korea. In addition, the study data collection process was standardized, and sample processing was centralized. However, the following limitations should be considered. First, the cross-sectional nature of the study precludes the imputation of causality for our findings. Second, sleep duration was self-reported and not determined by objective polysomnography, and thus, the responses could be subject to some degree of error. However, measurement error is likely to be non-differential with respect to recall of sleep duration and metabolic biomarker assessments. The study sample could have limited statistical power due to the small sample size; however, significant associations with triglycerides and cholesterol were identified. Lastly, although we adjusted for confounding factors, we cannot completely rule out the existence of potential residual or unmeasured confounding.

In conclusion, an increase in usual sleep duration was associated with elevated blood triglyceride and total cholesterol levels in Filipino immigrant women in Korea. Although sleep exposes complex relationships with the circadian rhythm, the underlying pathology underlying the association with metabolic risk factors still needs to be investigated for health promotion and disease prevention. Further prospective and experimental studies are warranted to understand the potential mechanisms that underlie the association between sleep duration and blood triglyceride and cholesterol levels.

Supplementary materials can be found online at https://doi.org/10.7570/jomes22032.

This study was financially supported by Hanmi Pharmaceutical Co., Ltd, (No. 201300000001270), Chong Kun Dang Pharm. Seoul, Korea (No. 201600000000225), and Handok Inc., Seoul, Republic of Korea. The funders had no role in study design, data collection or analysis, the decision to publish, or preparation of the manuscript. The content is the full responsibility of the authors and does not necessarily represent the official views of the funding agencies.

The authors are grateful to the participants of the study and the volunteer staff of FiLWHEL.

Study concept and design: GHC, SH, SHY, JEL, and CBL; acquisition of data: HSK, HL, SMPP, SH, SHY, JEL, and CBL; analysis and interpretation of data: HSK, HL, SMPP, and JEL; drafting of the manuscript: HSK, HL, and JEL; critical revision of the manuscript: HSK, HL, SMPP, GHC, SH, SHY, JEL, and CBL; statistical analysis: HSK; obtained funding: JEL and CBL; administrative, technical, or material support: GHC, SH, SHY, JEL, and CBL; and study supervision: JEL and CBL.