Study population
The data in this study are based on the the baseline survey of the Shanxi CIN Cohort Study in 2014, which includes 40,000 eligible local women from the Shanxi province, China. The rationale, design, and methods of the Shanxi CIN Cohort Study have been detailed elsewhere [21, 35], [36]. Briefly, we conducted a free cervical cancer screening for eligible women who were permanent residents of the two counties in Shanxi province during 2014.
A total of 40,000 participants were included. All participants were surveyed using a demographic characteristics-related questionnaire and a Pap test based on liquid-based cytology (LBC). Participants with abnormal Pap test results were examined by colposcopy and histopathology. A total of 2,769 women were diagnosed as having atypical squamous cells of undetermined significance (ASC-US) and above, and 78 were excluded (68 refused to participate and 10 showed glandular cell abnormalities). Of the 2,691 participants for whom pathologic results were available, 1,890 had negative results who with abnormal cytology with currently normal histology. Finally, 564 participants were histologically diagnosed with CIN grade 1 (CIN1), and 237 participants with CIN2 or above (CIN2+). Of the 1,890 women with negative pathologic results, 1,503 participants were included in the study analysis after the exclusion of 387 women who had not fully completed the three parts of the medical examination, including an in-person interview, a physical examination, and a clinical examination. A sample of 2,304 women with a mean age of 49.2±9.1 years was enrolled in the present study, Detailed flowchart of this study have been published elsewhere [36]. All inspections and detections were performed under double-blind conditions. The study was approved by the ethics committee of the Second Hospital, Shanxi Medical University and written informed consent was obtained from the participants.
Data collection
We collected several types of data in this study, including questionnaire-related data obtained from in-person interviews, and clinical data from physical examinations, laboratory tests and biospecimen collection. In-person interviews were conducted by trained interviewers using a standardized questionnaire. Demographic information included age; years of education; yearly income; tobacco smoking; age at menarche; menopause status; years of intrauterine device (IUD) use; and sexual activity in the menstrual period. Participants completed a 26-item food frequency questionnaire (FFQ). Clinical data were obtained from physical examinations and laboratory tests, including Pap tests, vaginal pH tests, and cervical biopsies. Data on IUD use, squamous-columnar junction (SCJ) visibility, had gynecologic surgery and vaginitis were also collected. Participants provided biological samples that were stored for future work (blood and cervical tissue specimens).
Clinical laboratory tests
All Pap tests were performed using the LBC method. At least two cytopathologists from the Second Hospital of Shanxi Medical University evaluated the cytologic results, which were reported using Bethesda System (TBS) 2001 terminology. All abnormal cytology slides were further reviewed for quality control by a senior cytopathologist who was blinded to the previous pathology results if the case was of type ASC-US+ or worse. Gynecology specialists from the Second Hospital of Shanxi Medical University identified patients with an abnormal cervix, and biopsy was performed by colposcopy (SLC-2000 device, Shenzhen Goldway Company) according to a standard protocol ≤12 wk after the Pap test. Gynecology specialists divided the cervix into quadrants and examined each quadrant. All visually abnormal areas were biopsied, and the quadrants without a visible lesion were biopsied at the SCJ (“random biopsy”). Endocervical curettage was also performed. The cases were classified as negative, CIN1, CIN2, CIN3, or squamous cell carcinoma (SCC). In the end, the pathologists performed a double-blinded observation of the Pap test results after diagnosis based on the cervical biopsy or endocervical curettage tissue specimens. If two pathologists presented different diagnoses, the samples were reviewed by another senior pathologist. The three pathologists reviewed difficult or equivocal cases together to arrive at a consensus on diagnosis.
HPV genotyping by HybriMax was performed using residual Pap test specimens with the HPV GenoArray Test Kit (HybriBio Ltd), and cases were divided into the high-risk HPV infection group and “others” (including low-risk HPV and negative cases). This assay can identify 21 types of HPV, including 15 high-risk types (16, 18, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, and 68). The low-risk HPV types include types 6, 11, 42, 43, 44, and CP8304, which were identified using the flow-through hybridization technique performed with a TC-96/G/H6 HPV DNA Amplification Analyzer and an HMM-2 fast nucleic acid molecule hybridization instrument (HybriBio Ltd).
Vaginal pH testing was performed with a pH test paper (Marok Darmstadt Germany) with the residual Pap test specimens (2304); vaginal pH values exceeding 4.5 were considered abnormal [37]. Cases were divided into two groups based on the pH results: normal pH group (>3.8, and< 4.5) and abnormal pH group (≥ 4.5).
Dietary nutrient assessment
The measurement of individuals’ dietary nutrients intake were calculated based on a food frequency questionnaire (FFQ) obtained from in-person interview. The FFQ in this study was nested in the standardized and structured epidemiological questionnaire of the Shanxin CIN Cohort study. The 24 h dietary recall dietary data were collected by trained interviews who recorded amounts of all the food items. The FFQ was designed based on the China Health and Nutrition Survey (CHNS) [38], [39]. Detailed descriptions of the dietary measurements have been published elsewhere [38], [40]. A total of 26 items food was included in the FFQ, which are main food sources for participants in this study. Based on a Chinese National Nutrition Survey in 2002, ten of 26 items food already account for about 85% of the total dietary intake in Chinese population. The FFQ included 26 food items: wheat flour, soybean, cabbage, egg, oats flour, bean curd, celery, cow milk, buckwheat flour, dried bean curd, spinach, pork liver, rice, broad bean, Chinese chives, sunflower seed, millet, potato, carrot, jujube, maize, mushroom, pumpkin, banana, liquor, and tea.
The FFQ data was analyzed using the US Department of Agriculture’s 1994–1996 Continuing Survey of Food Intakes [41]. The average food intake of individuals (gram/day) was calculated. Each nutrients intakes (grams/milligrams/micrograms per day) were then calculated by multiplying the daily food consumption amount (g per day) by the median nutrition content (g per 100 g/mg per 100 mg/μg per 100 μg of food) of that food. The nutrition values from all other FFQ items were combined to obtain the total daily nutrition values. However, we did not collect the dietary supplement information in this study because the prevalence of nutrients supplement use is very low in Chinese population [42]. We estimated the level of each dietary nutrient intake as the following equation: Total daily nutrition value = total amount of each food (g/mg/μg) / (day)* intake value per 100 g, per 100 mg, and per 100 μg. Total nutrition intake (g/mg/μg per day) was determined by calculating the sum of the daily nutrition values.
Assessing the test–retest reliability and relative validity of the FFQ.
We randomly selected 218 out of 2304 people as subjects. The study started from January 2019 and lasted for the subsequent six months. During the study period, three consecutive 24-hour studies (24-HRs) were conducted every three months. The first FFQ was administered during the first 24-HR. FFQ1 was administered during the second 24-HR, and FFQ2 was administered during the last 24-HR. The ‘weight estimation (WE)’ method for assessing food consumed for evaluation was estimated by the respondents for the weight of each food they consumed in the previous 24 hours [43]. The study design is shown in Supplemental Figure 1. Each participant was asked to provide the name and amount of food consumed during the previous 24 hours. If the previous day was a special day, for reasons such as banquets or travel, et al., we would record food consumption 24 hours ago, or choose another day to interview participants by telephone. Subjects were not informed of the results until the night before the interview. Record the amount of food mixed with a plate. According to the definition of food quality standard, recalled food is assigned to the corresponding food group.
For fruits consumed, the subjects were required to select the corresponding pictures representing the different sizes of each fruit and record the corresponding estimated weight (in grams). For commercial projects, such as sliced bread, cakes, packaged biscuits, pies and dumplings, record the unit weight (in grams) of the project. Trained interviewers manage FFQ and 24 hours through face-to-face interviews. Immediately check all records and resolve any ambiguities in the subject. During the whole study period, each participant had his own interviewer. Participants who did not satisfactorily complete the FFQs or missed more than one out of the four 24-HRs were excluded from the analyses. Subjects with implausible energy intakes (<500 Kcal or >5000 Kcal) were also excluded as described by previous studies. Extreme values were examined and excluded. A decision about whether or not to exclude the record from analyses was made according to the original FFQs and/or 24-HRs [44]. The validity of the FFQ methods was assessed by comparing the nutrient intakes derived from the FFQs . A 3-month FFQ was collected from the same subjects after administration of the first interviewer-administered FFQ.
Statistical analysis
Descriptive statistics were used to describe the frequency and proportion, and mean and standard deviation of the demographic characteristics. Participants’ characteristics were examined for significant differences with a Pearson’s chi-squared test for categoric variables. A logistic regression model was used to calculate the odds ratios (ORs) and their confidence intervals (CIs) for CIN risk in each nutritional ingredient quartile relative to that in the highest quartile. Tests for a linear trend across increasing quartiles of nutritional ingredients were performed by assigning the medians of each nutritional ingredient to quartiles treated as continuous variables.
Analyses were adjusted for potential confounders. The first model was unadjusted. Next, we adjusted for age (<30, 30-39, 40-49, 50-59, and >60 years), years of education (<6, 7–9, and >9 years), yearly income (<10,000, 10,000–30,000, and >30,000 ¥), tobacco smoking (yes and no), age at menarche (<13, 13-<15, 15-<17, and >17 years), menopause status (yes and no), IUD use (yes and no), years of IUD use (<10 and ≥10 years), sexual activity in the menstrual period (yes and no), history of undergoing a gynecologic surgery (yes and no), and presence of vaginitis (yes and no). In the final multivariable analysis, we added other potential clinical confounders, including high-risk HPV (positive and negative), SCJ visibility (fully visualized and not fully visualized), and vaginal pH (<4.5 and ≥4.5).
We performed cross-sectional analyses with three knots (25th, 50th, and 75th percentiles) of the 2,304 women to examine the association between log-transformed dietary intake levels and CIN risk. We did not explore the association between dietary intake and SCC risk because of the limited number of SCC cases (n = 19). Statistical analyses were performed using SAS software version 9.3. All reported P values were two-sided, with a significance level of 0.05.