Study purpose and design
The purpose of this study was to determine whether the odds of current joint pain and knee OA change according to prior breastfeeding status and duration. We conducted post hoc analysis of cross-sectional data on Korean women.
Study participants and examined variables
Participant selection
KNHANES is a sample database representing health and nutrition-related behaviors of Korean individuals nationwide. It is a statutory survey on the health behavior, chronic disease prevalence status, and food and nutrition intake status conducted by the Korea Centers for Disease Control and Prevention on a regular basis and is a government-designated statistics based on the Statistics Act of South Korea, which is actively utilized in various public health policies and academic research. As for the subjects of KNHANES, based on the latest data of Population and Housing Census of Korea each time, new demographic samples with the representability of Korean population aged 1 and over are selected, and the total participation rate of the survey is maintained at about 80%. The detailed survey guidelines and the raw data of KNHANES can be found in the website of KNHANES (http://knhanes.cdc.go.kr/)[17].
The number of subjects in the 5th KNHANES was 10,938 in 2010 and 10,589 in 2011, and the actual numbers of participants were 8,958 and 8,518, respectively. For this study, we included women aged ≥50 years who underwent OA radiography examination and answered the self-administered questionnaire on breastfeeding. Overall, 3,454 women were included (Figure 1).
Breastfeeding duration
Participants who answered “no” to “(adult) Do you have breastfeeding experience of one month or longer?” in the health survey were categorized into the non-breastfeeding group. In other words, both women with no experience of breastfeeding and those with less than 1 month of breastfeeding experience were classified as non-breastfeeding. If the response was “Yes,” in the “(adult) total breastfeeding duration” item, the subject was asked to directly fill in the blank column set as two digits for the year and month, respectively, in the units of one month. The subject would provide the response by adding the respective breastfeeding duration in case of breastfeeding of multiple children, and in the survey, the breastfeeding duration per child was not collected. The participants were further divided into 1–24, 25–48, and ≥49 months breastfeeding groups. In this study, we set the breastfeeding duration that exceeds the recommended duration of 2 years as long-term breastfeeding.
Joint pain
The survey assessed whether respondents experienced knee, hip, and lower back pain. Those who answered “yes” to “Have you experienced knee joint pain/hip joint pain/lumbar pain for longer than 30 days in the past 3 months?” were classified into the knee joint pain, hip joint pain, and lumbar pain groups. In the overall analysis, "joint pain" was defined as the case where there was at least one joint pain in the three sites.
Knee osteoarthritis
Radiography for OA diagnosis was conducted for participants aged ≥50 years. As for the radiography for knee joints, in standing position, the anteroposterior view of bilateral knee joints was taken and the bilateral lateral view with the knee joint bent about 30 degrees was taken. The radiological OA diagnostic values for the knee joint were classified based on the Kellgren-Lawrence Grading Scale, as follows: 0, normal; 1, suspicion of osteoarthritis; 2, mild osteoarthritis; 3, moderate osteoarthritis; and 4, severe osteoarthritis [18]. The final grade was determined in comprehensive consideration of the findings of both joints, and as a result of the readings by three specialists in the department of OA radiology in a university hospital. When two out of three assessors had the same grade, this grade was reported, and when all three assessors reported different grades, the primary assessor’s grade was used. Those diagnosed with mild, moderate, and severe OA were considered to have knee OA in our study.
Confounding variables
Sociodemographic variables
Numerical variables were used to express participant ages. Income level was divided into four categories (low, middle-low, middle-high, high) according to the average monthly equivalized household income (monthly household income/√number of household members).
Lifestyle variables
BMI (kg/m²) was used to divide participants into three groups: underweight (<18.5), normal (18.5–24.9), and overweight (≥25. Based on smoking status, participants were classified as non-smokers (never smoked), ex-smokers, and current smokers. Second-hand smoke exposure was defined as “exposure to indoor second-hand smoke at the workplace or in the house for one hour or longer per day.” Based on alcohol consumption, participants were divided into three groups: less than once a month, 1-4 times a month, and ≥5 times or more per month.
The level of daily activities was classified into resting, light activities, moderate or intense activities according to the amount of physical activities performed in the week before the survey. For the moderate physical activities, the subjects who had “performed moderate physical activities requiring slightly more effort than usual and slightly accelerating heart rate for more than 30 minutes each time for more than 5 days a week” were counted, and for intense physical activities, the subjects who had “performed intense physical activities requiring considerably more effort than usual and considerably accelerating heart rate for more than 30 minutes each time for more than 5 days a week” were counted. Those who “Conduct moderate or intense physical activities as well as walking for at least 30 minutes once a day for more than 5 days a week” were categorized into the moderate or intense activities group, and those who did neither were categorized into the resting group. The rest of the participants were included in the light activities group.
Medical history
Participants indicated their current dyslipidemia, diabetes, hypertension, and osteoporosis status. In case of diabetes, impaired fasting glucose levels were also included. Hypertension was defined as “the presence of systolic blood pressure of ≥140 mmHg or diastolic blood pressure of ≥90 mmHg or treatment with medication for hypertension.”
Obstetric and gynecological variables
The numbers of natural birth, cesarean section, and preterm birth were combined to obtain the total number of children, whereas the numbers of spontaneous and induced miscarriages were combined to obtain the number of miscarriages. For the question on menstruation, among the responses of 1. Before menstruation, 2. Yes, 3. No, 8. Non-applicable (under 10 years old), 9. Unknown, those who responded with “2. Yes” were defined as non-menopausal women and those responded with “3. No” or “9. Unknown” were classified as menopausal women. Women who took hormone supplementation were checked, and the groups were categorized into 1-11 and ≥12 months.
Data analysis
KNHANES uses stratified cluster sampling and weighted values. In this study, complex sampling design analysis was conducted using stratified, cluster, and weighted variables. All data analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA).
The participants’ sociodemographic, lifestyle, medical history, and obstetric and gynecological characteristics were analyzed according to breastfeeding duration. The complex samples general linear model and Rao-Scott chi-squared test were conducted to compare continuous and categorical variables among the non-breastfeeding (women with <1 month of breastfeeding experience) group and the 1–24, 25–48, and ≥49 months breastfeeding groups. The missing values of the corrected confounding variables were excluded in analysis.
Using the non-breastfeeding group as the reference group, logistic regression models adjusting for confounders were used to calculate the odds of having the outcome (joint pain or degenerative knee OA) for subjects who breastfed ≥1 month. In addition, the breastfeeding group was further classified into 1-24 months breastfeeding group, 25-48 months breastfeeding group, and ≥49 months breastfeeding group. Using the non-breastfeeding group as the reference group, logistic regression models were performed to calculate the odds of having the outcome for subjects in 1-24 months breastfeeding, 25-48 months group, and ≥49 months breastfeeding groups.
The confounding variables include age, BMI, household income, smoking, alcohol consumption, physical activity, diabetes, hypertension, children, abortion, menopausal status, hormone replacement therapy. The first regression model adjusted only for age (Model 1), and the second regression model adjusted for all confounding variables including age (Model 2). In addition, we calculated p for trend to determine whether OR for joint pain and degenerative knee OA prevalence significantly increased with an increase in breastfeeding duration.
As knee OA is largely affected by age [19], and the differences in mean age among the breastfeeding groups were fairly large, a sensitivity analysis was conducted after re-selecting the participants aged ≥60. The OR and 95% CIs of knee OA were calculated in groups aged ≥60 years after adjusting for the same confounding variables. In all tests, p<0.05 indicated statistical significance.