2.1. Study population
The present study used data from the 2018 wave of CLHLS, a longitudinal population study initiated in 1998 with follow-up surveys every 2 to 3 years. The CLHLS surveys were conducted in randomly selected counties and cities in China, which accounted for half of the counties and cities in 23 out of 31 provinces covering over 85% of China’s population. Details of this survey have been published elsewhere 31, 32. The CLHLS is a specially designed sample with oversampled centenarians and very old adults aged in 90 s and 80 s.
In the 2018 wave of CLHLS, the self-reported types and frequencies of tea intake and depressive symptoms assessed by the Center for Epidemiologic Studies Depression Scale were collected. After excluding 2,469 participants with missing data on depressive symptoms, self-reported types of tea consumption, key covariables, the final analytical sample included 13,115 participants aged over 65 years old (5,121 were aged 65–79, 6,301 were aged 80–99, and 1,693 were aged over 100 years old) (Fig. 1).
The CLHLS study was approved by the Biomedical Ethics Committee of Peking University (IRB00001052-13074). All participants or their legal representatives signed written consent forms in the baseline and follow-up surveys.
2.2. Measurements
The questionnaire in the 2018 wave of the CLHLS included items about the frequency of habitual consumption of 8 types of tea (green, black, Oolong, white, yellow, dark, compressed, and flowering teas). The detailed types and classifications of tea consumption in this study are provided in supplements (Supplementary Table 1). In brief, we classified the type of tea into Green tea, Fermented tea (black, Oolong, white, yellow, dark, and compressed teas), and Flower tea 33. We grouped the frequency of tea consumption of each type of tea into 3 categories: daily (≥ 1 cup/day), occasionally (< 1 cup/day but ≥ 1 cup/month), and never or rarely (< 1 cup/month or never drink tea) 34.
We used the 10-item of the Center for Epidemiologic Studies Depression Scale (CES-D-10) to measure depressive symptoms in this study 35. The answers are indicated in a four-scale metric, from “rarely” to “some days” (1–2 days), “occasionally” (3–4 days), or “most of the time” (5–7 days). For the two positive questions— “I was happy” and “I felt hopeful about the future”—answers were reversely coded before summation. We then coded all answers from 0 to 3 as “rarely” and “most of the time”, respectively. The total range of CES-D-10 scores in this study was 0–30, with higher scores indicating greater severity of depressive symptoms. A person is considered to have depressive symptoms if he/she scored less than 10 in the CES-D-10. This threshold of 10 has been widely used in previous studies 36 and well validated in depression measurement in Chinese older populations, regardless of their age and dementia status 37, 38.
The 8th survey of CLHLS in 2018 collected a range of self-reported data on demographic, socioeconomic, psychosocial, behavioral, health-related factors, including age, gender, education, socioeconomic level, rural residence, geographical regions, marital status, living condition, social and leisure activity index, smoking, alcohol drinking, BMI, regular dietary (vegetable/fruit/fish/nut) intake, self-rated health, cognitive impairment, medical illness, comorbidity, and ADL disability. All information was collected through face-to-face home interview by trained research staff members. Interviewees were encouraged to answer as many questions as possible. If they were unable to answer questions, a close family member or another proxy, such as a primary caregiver, provided the answers 39.
Age was calculated according to self-reported dates of birth. If dates were converted into Georgian calendar dates if they were based on Chinese lunar calendar dates. Levels of educational attainment were as grouped into three categories according to years of schooling (0, 1–6, and ≥ 7 years). Marital status was divided as “currently married and living with spouse” or others (widowed, separated, divorced, or never married). Living condition was grouped into 3 categories: living with family members or others, living alone, and living in an institution. Current residence was dichotomized as “urban residence” or “rural residence”. Smoking status was dichotomized as “non-current smoker or never-smoker” vs. “current smoker”, a similar approach was taken to define the alcohol consumption and physical activity. Dietary intake, included vegetables, fruit, fish, and nut, were dichotomized as “regular intake” or “occasional or seldom intake”. The body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Given that no direct indicator for individual socioeconomic status is provided in the CLHLS, we obtained individually socioeconomic status by using a principal component analysis (PCA) based on four questions (primary occupation before retirement [white collar vs. others], living conditions [living with family members or others, living alone, and living in an institution], retirement earnings, and living expenditure). A compositional score based on the first component generated from PCA has been suggested to be a qualified measure of socioeconomic status and has been widely employed in previous studies 36, 40. Social and leisure activity score was calculated by eight kinds of activities (whether a respondent did gardening, practiced Tai Chi, participated in square dance, raised poultry or pets, reading, playing Mahjong or cards, listening to the radio or watching TV, and participating in community social activities) and we scored each activity 1 for ‘never’, 2 for ‘sometimes’ 3 for ‘almost every day’; The score ranged from 8 to 24 with higher score indicating more leisure activities, and low social and leisure activity level was defined by the score less than 14. Cognitive function was tested by using the Chinese version of the 30-point Mini-Mental State Examination (MMSE) and cognitive impairment was defined by an MMSE total score of < 24 32. Activity of daily living (ADL) was assessed by the Katz index 41, we defined ADL disability as needing personal assistance in performing one or more of the five essential activities (bathing, transferring, dressing, eating, and toileting) or being incontinent 42. We ascertained 14 self-reported medical illnesses, including hypertension, diabetes, dyslipidemia, heart disease, stroke, pneumonia (asthma/COPD), cataract or glaucoma, cancer, gastritis, arthritis, cholecystitis, rheumatism, nephritis, and hepatitis; we grouped the medical illness into 3 categories: “chronic inflammatory disorders (heart disease, stroke, diabetes, pneumonia, gastritis, arthritis, cholecystitis, rheumatism, nephritis, and hepatitis)”, “other disorders”, and “none”. Comorbidity was defined as having 5 or more medical illnesses. Self-rated health was defined as “excellent or good” or “average or poor”. We considered geographical region on the basis of residential address to account for types of tea production areas 43 as well as differences in regional economic developments and social cultures in China: Northern China (Beijing, Tianjin, Hebei, Shanxi, Shaanxi, Shandong, Liaoning, Jilin, and Heilongjiang provinces), Eastern China (Shanghai, Jiangsu, Zhejiang, and Fujian provinces), Central China (Henan, Hubei, Jiangxi, Anhui, and Hunan provinces), Southwestern China (Guangdong, Guangxi, Chongqing, Sichuan, and Hainan provinces) (Supplementary Fig. 1).
2.3. Statistical Analyses
The subjects’ characteristics according to categories of type of tea consumption were compared by using analysis of variance or chi-squared test, as appropriate. We used multivariate logistic regression analysis to calculate odds ratios (ORs) for depressive symptoms relative to the type of tea consumption including green tea, fermented tea, and flower tea, with no habitual tea intake treated as the reference group. The base model (Model 1) included types of tea consumption plus demographic variables; Model 2 further controlled for socioeconomic variables: education, socioeconomic status, rural residence and geographical regions; Model 3 additionally controlling for psychosocial and behavioral variables: marital status, living condition, social and leisure activity index, smoking, alcohol drinking, BMI, regular dietary (vegetable/fruit/fish/nut) intake; Model 4 added health variables in Model 3: self-rated health, cognitive impairment, medical illness, comorbidity, and ADL disability. In detailed analyses examining the dose-effect relation between the intake of green tea or fermented tea or flower tea with depressive symptoms, we classified the frequency of each type of tea consumption into 3 categories: daily (≥ 1 cup/day), occasionally (< 1 cup/day but ≥ 1 cup/month), and never or rarely (< 1 cup/month or never drink tea), and repeated multiple logistic regressions controlling for all covariates as above.
We conducted subgroup analyses to examine whether the associations between types and frequencies of tea intake and depressive symptoms differed by gender, age (< 80 years old vs. ≥80 years old), residence (urban residence vs. rural residence), and geographical regions (Northern China, Eastern China, Central China, and Southwestern China). We performed several steps of sensitivity analyses for the full model (Model 4) to assess the possible outcomes of the different thresholds used for the CES-D-10. First, we considered varied cut-off thresholds for the CES-D-10, such as 8 and 12, which are more sensitive (cut-off value = 8) or specific (cut-off value = 12) to discriminate the depressive symptoms, and used the Model 4 to examine the associations. Second, we excluded the participants with severe cognitive impairment with scores of MMSE < 19 44, of whom substantial recall bias might have occurred in reporting types and frequencies of tea consumption. Moreover, we removed the participants who were long bedridden or terminally ill, restricting the sample to non-bedridden to see whether there is a change in the significance level of the observed associations. We also test our results by using full sample after multiple imputation and by adjusting sampling weight based on age-sex-residence-specific distribution of 2015 mini-census of China.
A two-tailed P-value of less than 0.05 was considered statistically significant. All analyses were performed using STATA version 14.0 (Stata Corp, College Station, TX, USA). ArcGIS version 12.0 was used to perform map visualization of the geographical distribution of tea drinkers.