From April 2001 to May 2001, a large cross-sectional, community-based epidemiological study was conducted. The subjects for the survey were adults aged between 35 and 59 years. A two-stage cluster-sampling scheme based on existing census divisions was used to randomly select four areas, each with a population from 300 to 350 subjects, and sample was stratified by sex and age group (5 years) to ensure representation of each part of the population. Among the 1 351 individuals investigated, the response rate was 88.5%, and the random sample and random-sample responder populations closely reflected the actual distribution of age group and sex in Pizhou area. Among 1 193 rural residents, there were 530 males (44.57%) and 663 females (55.43%), and their average age was 46.69 years. Signed informed consent was obtained from all participants and the study was approved by the Nanjing Medical University Ethics Review Committee.

A standard questionnaire was adopted, and the investigators were students from Nanjing Medical University, who had received special training. The questionnaire included questions about occupation, height, weight, sleep habits, sleep quality, and frequency of disruptive snoring. The question concerning snoring was “How often do you snore loudly and disturbingly?” with the following response alternatives: 1 = never, 2 = seldom, 3 = sometimes, 4 = often, 5 = very often, and those who answered 2-5 were regarded as snorers.

Anthropometric measurements were performed after the participants had removed their shoes and upper garments and donned an examining gown. Each measurement was performed twice and the average was used in the analysis. Height was measured to the nearest 0.1 cm using a wall-mounted stadiometer. Weight was measured to the nearest 0.1 kg using a hospital balance beam scale. Body mass index (BMI) was calculated as weight (kg) divided by the square of height (m²). The waist circumference was measured to the nearest 0.5 cm at the point of narrowing between the umbilicus and xiphoid process (as viewed from behind) and the waist circumference was used as a judgment of upper-body adioposity. Blood pressure was measured in the right arm with the participant seated and the arm bared. Three readings were recorded for each individual, and the average of the three readings was defined as the subject’s blood pressure.

The 12-h fasting blood samples were drawn in the morning and the sera were stored at -70 °C immediately after centrif-ugation until being assayed. All laboratory measurements were conducted at the Central Clinical Laboratory in the First Affiliated Hospital of Nanjing Medical University. Fasting blood glucose (FBG), fasting total cholesterol (TCH), fasting triglyceride (TG) and fasting high-density lipoprotein cholesterol (HDL-c) were determined by enzymatic procedures on an automated autoanalyzer (AU 2700 Olympus, 1st Chemical Ltd., Japan), and the source of reagents for the autoanalyzer was from 1^st Chemical Ltd., Japan. The laboratory was monitored for precision and accuracy of glucose and lipid measurements by the agency’s surveillance program. Measurements of agency-assigned quality control samples showed no consistent bias over time within or between surveys. Low-density lipoprotein cholesterol (LDL-c) was assessed by the Friedwald method[7], and the concentration of LDL-c was calculated for each subject according to the following formula: C_LDL-c = C_PLASMA-C_HDL-TG/5.

The intact insulin level was measured using a highly sensitive two-site sandwich ELISA[8]. This assay method for insulin is based on the enzyme immunoassay principle and it is constructed as a sandwich immunoassay. Two monoclonal murine antibodies specific for insulin were employed. One of the antibodies (HUI-018), which binds to an epitope on one side of the insulin-molecule, was used for coating on the ELISA plate wells. The other antibody (OXI-005), which binds to another epitope on the other side of the insulin-molecule, was covalently bound to biotin. The biotin-antibody reagent and sample or calibrator were added to the precoated-wells followed by incubation. During incubation, a complex between plastic-surface, coated-antibody, insulin and the biotinylated-antibody was formed. This complex would not be removed by the washing procedure, which followed the incubation. After the wash, avidin-peroxidase was added. It bound to the biotin of the biotinylated-antibody and extended the complex with the enzyme peroxidase. The amount of enzyme was visualized by addition of the substrate peroxide and 3’,3’,5’,5’-tetramethylbenzidine (TMB), which was converted to a soluble colored product. The developed color was proportional to the amount of insulin in the sample. The color-development progressed with time and it was interrupted after a fixed time by the addition of phosphoric acid. The color was stable and absorbance was measured in an ELISA-plate photometer. A calibrator curve was constructed based on the absorbance values and the insulin concentration in the serum samples could be found. The detection limit was 5.0 pmol/L. The specificity of the assay excluded intact, split (32-33) and des (31, 32) PI. There was some cross-reactivity with the less abundant split (65-66) PI (30%) and des (64, 65) PI (63%).

The PI level was measured in a similar manner using another sensitive two-site sandwich ELISA[9]. The analysis for human PI is an enzyme immunoassay constructed as a sandwich immunoassay. Two monoclonal mouse antibodies were used: one with specificity for human C-peptide (PEP-001) was used for coating, the other antibody (HUI-001) with specificity for insulin was biotin labeled and acted as detecting antibody. The plasma samples were diluted 1:2 during analysis to eliminate the plasma effect. In the first reaction, after the samples were applied to the coated wells, PI in the samples would bind to the coating antibody. In the second reaction, the bound PI was detected by the use of another mAb directed against human insulin and biotin labeled. In a third reaction, the amount of bound detecting antibody in the coating antibody-PI-detection antibody complex was visualized by the use of streptavidin-peroxidase that bound to the biotin label. Between each step, the plate was washed four times in a washing buffer in order to measure only the bound material. Finally, a substrate for the peroxidase (tetramethylbenzidine and H₂O₂) was applied to the wells resulting in color development proportional to the amount of peroxidase, i.e., PI present. The enzymatic reaction resulted in development of a blue color, which changed into yellow, when the reaction was stopped with H3PO4. The plate was read in an ELISA reader at 450 nm and the absorbance at 620 nm was subtracted. The detection limit in human serum was 0.25 pmol/L. There was no cross-reactivity with human insulin and human C-peptide. However, the four major PI conversion intermediates reacted at various proportions of 65-99%.

The between- and within-assay coefficients of variation were 6.8, 7.8% for true insulin respectively and 6.7, 7.8% for PI respectively. All measurements were performed in duplicate. The four mAbs including OXI-005, HUI-018, PEP-001 and HUI-001 were kind gifts from Novo Nordisk, Bagsvaerd, Denmark, and the standard samples of true insulin and PI were supplied by the Mercodia Company, Sweden. The ELISA-plate photometer (Bio-tek EL900, USA) and ELISA plate (NUNC company, Denmark) were employed in the assay.

To investigate the relationship between the level of PI and cardiovascular risk factors clustering, we created a risk factor score to rank individuals according to the number of the risk factors at the time of the survey. The following risk factors and cut-off points were used to build up the risk factor scores: (1) hypertension was defined when systolic blood pressure (SBP) was ≥140 mmHg and/or diastolic blood pressure (DBP) ≥90 mmHg or taking anti-hypertensive drugs because of previous hypertension according to the 1999 WHO/ISH criteria[10]; (2) hyperglycemia was diagnosed based on the FBG serum glucose ≥6.1 mmol/L according to the American Diabetes Association (ADA) criteria[11] or having a history of diabetes mellitus; (3) hypercholesterolemia was defined as fasting TCH≥5.20 mmol/L; (4) high LDL-c was defined as LDL-c≥3.38 mmol/L; (5) low HDL-c was defined as HDL-c≤1.04 mmol/L; (6) hypertriglyceridemia was defined as fasting TG≥1.70 mmol/L[12]; (7) overall overweight was considered as BMI ≥25.0 kg/m2 according to the WHO guideline[13]; (8) visceral obesity was defined as waist circumference ≥85 cm in male and ≥80 cm in female[14]. The final risk factor score varied from 0 to 5, with 0 meaning no exposure to these risk factors; one exposure to any one risk factor; 2-4 exposure to any combination of 2-4 risk factors respectively; five exposure to any combination of five or more than five risk factors simultaneously.

All data analyses were performed using statistical package for social science (SPSS for Windows, version 10.0, 1999, SPSS Inc, Chicago, IL, USA). Data of BMI, waist circumference, age and blood pressure were normally distributed parameters and presented as mean±SD, whereas skewed data including FBG, fasting lipid, fasting true insulin, and PI were logarithmically transformed before analysis and were expressed as median (quartile range).

Tables 1 and 2 show sex and age distribution of sleep snoring in the population-based epidemiological study. The results indicated that the frequency of sleep snoring in males was significantly higher than in females (χ² = 68.227, P<0.001), and the frequency of sleep snoring significantly increased with the elevation of age (χ² = 44.465, P<0.001).

The results of Spearman correlation analysis (Table 3) indicated that the frequency of sleep snoring was significantly correlated with such parameters as BMI, waist circumference, SBP, DBP, fasting TCH, fasting TG, fasting LDL-c, PI, and risk factors score (P<0.01). However, the significant correlation was not found between the frequency of sleep snoring and FBG, true insulin, and fasting HDL-c (P>0.05).

Table 4 summarizes the results of analysis of covariate variance of sleep snoring and its relevant parameters. In the analysis, age and sex were controlling variables; the frequency of sleep snoring was factor variable, and the parameters including BMI, waist circumference, SBP, DBP, FBG (logarithmically transformed value), TCH (logarithmically transformed value), TG (logarithmically transformed value), HDL (logarithmically transformed value), LDL (logarithmically transformed value), true insulin (logarithmically transformed value), PI (logarithmically transformed value), and risk factors score were employed as dependent variables, respectively. The results indicated that a significant difference in parameters including BMI, waist circumference, SBP, DBP, TCH, TG, true insulin, PI, and risk factors score was found among various sleep snoring groups respectively after controlling for age and sex, and significant difference was not found in FBG, HDL-c, LDL-c between different frequencies of sleep snoring groups.

The results of multivariate logistic regression analysis are shown in Table 5, in which the dependent variable was the frequency of sleep snoring (the never snoring was coded as 0, and the seldom, sometimes, often, and very often snoring was considered as sleep snoring and was coded as (1), and age stratum (age stratification was performed with 5-year age strata), sex (male was coded as 1, female as 0), BMI (group division by an interval of quartile), waist circumference (group division by an interval of quartile), true insulin (group division by an interval of quartile), and PI (group division by an interval of quartile) were employed as independent variables. In the analysis of multivariate logistic regression, the variables including sex, age, PI, and waist circumference were accepted by the final model, and BMI and true insulin were refused by the regression model. The final model suggested that male, age, PI, and waist circumference were risk factors for sleep snoring, and the OR (95%CI) for sex, age, PI, and waist circumference was 0.401 (0.311-0.518) (P = 0.000), 1.249 (1.146-1.361, P = 0.000), 1.220 (1.085-1.373) (P = 0.001), and 1.398 (1.239-1.578) (P = 0.000), respectively.