Data were obtained from the Hallym Aging study (HAS). HAS is a prospective cohort study of health among elderly community-dwelling residents of Chuncheon, a city about 120 km east of Seoul. This ongoing study began in 2004, with follow-up examinations planned every 3 yr. The methods have been described elsewhere in detail (8). Briefly, eligibility criteria include age of 50 yr or older and residence within the borders of the survey area for at least 6 months before the survey. Using Korean National Census data for the year 2000, 200 of 1,408 census tracts were randomly sampled according to residential area (9). Study subjects were selected so that those over 65 yr old represented about 70% of the study cohort. This study involved subjects from the second triennial examination. Of the 702 eligible participants, 129 subjects declined knee radiographs and 69 subjects were excluded due to poor knee radiograph quality (n = 4), total knee replacement surgery (n = 3), or missing films due to clerical error (n = 62). Those who declined to take radiographs were significantly older than the subjects who participated. The remaining 504 subjects were analyzed for this study.

Demographic information was collected using a standard questionnaire, which included educational attainment, marital status, household income, occupation, and comorbidities. Work demanding physical exertion (e.g., sometimes carrying heavy objects, using instrumentation) or heavy physical exertion (e.g., construction worker, laborer, farmer) were defined as a manual occupation. Comorbidity health information was self-reported according to a closed questionnaire listing 29 pre-defined illnesses. Height (cm) and body weight (kg) were measured to the nearest 0.1 cm and 0.1 kg, respectively, with the subject wearing light clothing and barefooted for calculation of the body mass index (BMI).

Knee pain was assessed by asking, "Have you had pain, aching, or stiffness lasting at least a month in your knee?" All subjects also completed the Korean version of Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), a cross-culturally adapted and validated instrument that measures knee pain and physical function in persons with knee OA (10), and the Short Form 12-item (SF-12) questionnaire for the evaluation of self-reported functional status and QOL.

Physical performance of lower extremity was tested using the Health-ABC battery, as reported previously, with modifications (11). Briefly, standing balance, a 6-m usual walk, and five repeated chair stands were tested. To test standing balance, the subjects were asked to attempt to keep their feet in side-by-side, semitandem, and tandem positions for 10 sec each. The subjects were given a score of 0 if they could hold a side-by-side standing position for 10 sec but were unable to hold a semi-tandem position for 10 sec, a score of 1 if they could hold a semi-tandem position for 10 sec but were unable to hold a full tandem position for more than 2 sec, a score of 2 if they could stand in the full tandem position for 3 to 9 sec, and a score of 3 if they could stand in the full tandem position for 10 sec. For the 6-m usual walk test, a 6-m walk at the subject's normal pace was timed twice, and the time of the faster of the two walks was used for scoring. For repeated chair stands, the subjects were asked to fold their arms across their chests and to stand up and sit down five times as quickly as possible, and the length of time required was measured. For repeated chair stands and 6-m usual walks, we created quartiles based on the performance of the study subjects, with quartile 0 representing the worst and quartile 3 representing the best performance. Research nurses trained in lower extremity performance test carried out the assessments. To avoid inter-assessor variability, one nurse took charge of one item of the test of all participants. Previously, test-retest correlations of more than 0.89 for walking speed (12), 0.73 for chair stand (13) and 0.97 for standing balance (14) have been reported for these measures.

Radiographic evaluations consisted of weight-bearing anteroposterior, 14 × 17-inch, semi-flexed knee radiographs. Each knee was graded using the Kellgren and Lawrence (K-L) grade (15). Radiographic knee OA was defined as being present if the subject had a radiographic grade in the tibiofemoral joint of ≥ K-L grade 2. Radiographs were read twice by one reader, an academically based rheumatologist. Films allocated different K-L grades at the two readings were adjudicated by consensus between the original reader and a second reader, another academically based rheumatologist. The reproducibility of intra-reader assessments was high (for OA vs no OA, κ = 0.89).

Subjects were divided into four age groups: 50-59 (n = 56), 60-69 (n = 126), 70-79 (n = 274), and ≥ 80 yr (n = 48). The age-specific prevalence of knee pain was calculated for men and women. Student's t test and chi-square tests were used to compare the subjects with or without knee pain. Logistic regression models with generalized estimating equations (GEE) were used to calculate the odds ratios (ORs) for the presence of knee pain. GEE were used to minimize the lack of independence between measures originating from the same individual. In multivariate GEE analysis, an adjusted OR was calculated after adjusting for age and factors found to be significant in univariate analysis. Because a sizable majority of subjects marked 0 for WOMAC subscales, the subscales were categorized into quartiles 0-3 with quartile 3 representing the worst scores. The ORs and 95% confidence interval (CI) for belonging to the worst quartile of WOMAC were calculated by logistic regression analysis in subjects with knee pain compared to subjects without knee pain after adjusting for age, BMI, sex, and the presence of knee OA. In addition, sex-specific quartiles were created and analyzed to separately compare male and female subjects, respectively. Scores for SF-12 items were analyzed and compared using general linear models after adjustment for age, sex, BMI, and the presence of knee OA between subjects with and without knee pain. The proportion of subjects belonging to each category of physical performance test was compared between subjects with and without knee pain using chi-square test. Data were analyzed using SAS software version 9.1 (SAS Institute, Cary, NC, USA). P values < 0.05 (2-tailed) were considered statistically significant.

This study was approved by the institutional review board of the Hallym University School of Medicine (IRB approval number: HIRB-2007-001). All participants agreed to the use of their health check-up results including blood sample, physical performance test and knee radiographs. Written informed consent was obtained from each participant.

Of the 504 participants, 54% were women and the mean age was 70.2 yr. The demographic and clinical characteristics of the study population are shown in Table 1.

Those with knee pain were more likely to be female. The subjects with knee pain had a significantly lower level of education, lower income, higher BMI, and higher frequency of manual occupation and living without a spouse. Of the subjects with knee pain, 55.8% had radiographic knee OA.

The overall prevalence of knee pain was 46.2% (32.2% in men and 58.0% in women, P < 0.001). In 10.3%, 9.1% and 26.8% of subjects, pain was present in right, left or both knees, respectively. The prevalence of unilateral knee pain in dominant leg was 10.4% and 16.1% in right and left knees, respectively. Fig. 1 shows the age-specific prevalence of knee pain in men and women. Except for subjects under the age of 60 yr, the prevalence of knee pain was significantly higher in women than in men. The prevalence increased with age in women until 70 yr then leveled off (P < 0.001 for trend). The prevalence of knee pain did not increase with age in men, and the mean age of subjects with knee pain was not significantly different from that of subjects without knee pain (Table 1).

In univariate analysis, knee pain was associated with female gender, higher BMI, lower educational level, non-smoking, manual occupation, living without a spouse, lower income, and the presence of radiographic knee OA (Table 2). None of these variables had significant correlations with each other (Pearson correlation < 0.5, data not shown), thus, multivariate analysis was performed after adjusting for age and all of those significant factors. After multivariate GEE analysis, female gender (OR = 2.82, 95% CI = 1.47-5.40) and the presence of knee OA (OR = 2.81, 95% CI = 2.02-3.89) were significantly associated with knee pain. Knee OA was significantly associated with knee pain in both genders with adjusted ORs of 2.44 (95% CI = 1.31-4.53) and 2.92 (95% CI = 1.96-4.36) in men and women, respectively. Among subjects with knee OA, 47.2% of men and 70.2% of women had knee pain, while among those without knee OA, 29.4% of men and 40.7% of women had knee pain. Women had significantly higher ORs for knee pain compared to men in both radiographic OA and non-OA groups (adjusted ORs, 3.22 [95% CI = 1.66-6.24] and 1.94 [95% CI = 1.28-2.96], respectively).

To compare the pain, stiffness, and function as measured with WOMAC between subjects with and without knee pain, we performed logistic regression analysis to estimate ORs for belonging to the worst quartile (Table 3). After adjusting for age, sex, BMI and the presence of knee OA, subjects with knee pain had significantly increased risk of belonging to the worst pain, stiffness, and function quartile compared to subjects without knee pain. In both men and women, knee pain was correlated with worse WOMAC scores compared to those without knee pain. Among subjects with knee pain, women had significantly higher risk for belonging to the worst quartile for all WOMAC subscales compared to men after adjustment of age, BMI, and the presence of knee OA.

The mean scores for the SF-12 items are shown in Table 4. Similar to the WOMAC results, subjects with knee pain had significantly worse (lower) scores on all SF-12 subscales than did subjects without knee pain after adjusting for age, sex, BMI, and the presence of knee OA in both genders. Among those with knee pain, women had worse SF-12 scores in all categories except for general health perception and social functioning in comparison to men.

We compared the proportion of subjects belonging to each category of Health-ABC battery according to the status of knee pain (Table 5). The subjects with knee pain had worse scores on all tests than subjects without knee pain, and knee pain was significantly associated with poorer physical performance. While female subjects with knee pain had significantly worse functional categories compared to those without knee pain in all tests, male subjects with knee pain did worse in usual walk test only. Although standing balance was not influenced by gender, men performed significantly better on usual walk (mean 6.0 vs 6.9 sec in men vs women), and repeated chair stands (mean 11.91 vs 13.40 sec in men vs women). Regarding this as inherent gender difference, we did not compare the lower extremity function between men with OA and women with OA.