Trends and disparities in osteoarthritis prevalence among US adults, 2005–2018

Studies reporting trends and disparities of osteoarthritis (OA) in the United States are limited. We aimed to examine trends and disparities of OA prevalence among US adults, from 2005 to 2018. Continuous National Health and Nutrition Examination Survey (NHANES) data from 2005–2006 to 2017–2018 were analyzed. Age-adjusted and self-reported OA prevalence, stratified by race/ethnicity and socioeconomic status (SES), was calculated separately for men and women. The linear trend and the association between the survey cycles and OA prevalence were assessed. Age-adjusted and self-reported OA prevalence linearly increased in the seven survey cycles (both Plinear trend ≤ 0.0002) in men and women. Non-Hispanic Caucasians (both Plinear trend ≤ 0.0001) in both genders and Non-Hispanic African Americans women (Plinear trend ≤ 0.0001) had significantly increasing linear trends in OA prevalence. In addition, people with lower SES had a lower age-adjusted prevalence of self-reported OA when compared to those with higher SES. The increasing linear trends still existed among both men and women after adjusting for multiple confounders (both Plinear trend ≤ 0.002). There were significant rising trends and disparities in self-reported OA prevalence among US men and women between 2005 and 2018.

www.nature.com/scientificreports/ examination includes medical, dental, and physiological measurements; the detailed methodology and protocols have also been described elsewhere 12 . To produce reliable statistics, NHANES oversamples persons 60 and older who are of African American and Hispanic ethnicity. NHANES study protocol has been approved by the National Center for Health Statistics Research Ethics Review Board. Written informed consent was obtained for all adult participants. All research was performed in accordance with the Declaration of Helsinki. Sample weights in NHANES have been constructed to adjust for non-response, oversampling, and non-coverage. Because of the thoroughness of its research methodology, NHANES data have been widely used over the years to reliably assess many diseases' prevalence and risk factors. NHANES only collects osteoarthritis information among adults aged 20 or older, which consists of the analytic population in the current study.
Variables. Since self-reported, doctor-diagnosed arthritis is the most commonly used case definition for prevalence and other epidemiological studies [13][14][15] . Each NAHNES participant was defined as having OA if he/ she answered: "yes" to the question "Has a doctor or other health professional ever told you that you had arthritis?" and "osteoarthritis" to the question "Which type of arthritis was it?" Demographic variables, including age, gender, and race/ethnicity, were ascertained by questionnaire. For the race/ethnicity groups, Mexican American and Other Hispanic were merged into Hispanic, and the remaining groups were Non-Hispanic Caucasian, Non-Hispanic African American, and Non-Hispanic Other, respectively. The educational attainment and family poverty income ratio (PIR) of participants were chosen as SES indicators. Educational attainment was categorized as less than high school, high school graduate/GED, some college, and college graduate or above 16 . PIR was computed as a ratio of the mid-point of the observed family-income category to the family's appropriate poverty threshold in a given calendar year, as set by the US Census Bureau 17 . Individuals were stratified into three levels based on their PIR: PIR < 1.3 (low income), 1.3 ≤ PIR < 3.5 (middle income), and ≥ 3.5 (high income) 18 . The cutoff point for participating in the Supplemental Nutrition Assistance Program is PIR = 1.3, so individuals with PIR < 1.3 were classified to the low-income group; PIR ≥ 3.5 provides relatively equal sample sizes for each of the three income groups 18 , thus people with PIR ≥ 3.5 were classified to high-income group and those with 1.3 ≤ PIR < 3.5 were the middle-income group. OA-related risk factors were considered and selected based on existing literature and on availability in the NHANES data. Weight status 19 , smoking status 20 , and physical activity 21 were included in the current study. Participants were categorized as obese if body mass index (BMIweight in kilograms divided by height in meters squared) was greater than 30 22,23 . Smoking status was categorized into current smokers, former smokers, and non-smokers 24 . Current smokers were respondents who had smoked at least 100 cigarettes during their lifetime and reported smoking either every day or some days at the time of the interview. Former smokers were those who reported smoking 100 cigarettes during their lifetime but currently did not smoke. Otherwise, participants were classified as non-current-smokers. Physical activity was categorized as inactive and active. Participants who were sedentary or only did basic activities, which refers to the light-intensity activities like standing and walking slowly, were considered to be inactive; otherwise, the individuals were classified as active 25 .
Statistical analyses. Sampling weight was used to account for the complex survey design (e.g., unequal probabilities of selection) during analysis. Estimates were age-adjusted by the direct method to the 2000 US Census population 26 . Age-adjusted OA prevalence in every survey cycle was estimated by race/ethnicity, education level, and PIR level for each gender. Standard errors, which were employed to construct confidence intervals, were estimated using Taylor series linearization. Testing for a difference of age-adjusted prevalence between groups was done using the pairwise t-test. Linear trends during the seven survey cycles were assessed by gender, race, and SES using orthogonal polynomial contrasts. JoinPoint Software (National Cancer Institute, Bethesda, MD) was utilized to determine the slopes and find the inflection point and differences in slopes between the two survey cycles by using piecewise linear regression 27 . If at least one significant change point was found, we report the year the trend shifted; otherwise, we only report P for linear trend. The survey cycle was used as a categorical variable in the analysis. OA prevalence was modeled as a function of the survey cycle after first adjusting for age and then with further adjustments for age, race, educational attainment, and PIR. We performed the analysis using the completed data, and missing data were excluded from the study. Since all variables had < 10% missing data, using complete data is unlikely to cause a biased estimate. Data analysis was conducted using procedure PROC SURVEY of SAS 9.4 (SAS Institute, Cary, NC, USA).

Results
Characteristics of the analytic sample. A  OA prevalence by race in both genders. The age-adjusted self-reported OA prevalence by race/ethnicity in men and women is presented in Fig. 2. Non-Hispanic Caucasian men had a higher age-adjusted prevalence of OA than men from the Hispanic and Non-Hispanic African American groups (both P-values < 0.0001). In comparison, Non-Hispanic women had a significantly higher OA prevalence than women in the other three Table 1.    OA prevalence by SES in both genders. The pattern of the age-adjusted prevalence of self-reported OA, stratified by education attainment in both men and women, is presented in Fig. 3. For men, the highest education level (≥ college) had a higher prevalence than other groups (all P-values ≤ 0.01). However, significant increasing linear trend were only observed among men with high school diploma/GED (P linear trend = 0.01) in 2005-2018 www.nature.com/scientificreports/ (slope = 0.65, P-value = 0.04). In men with high school diploma/GED, the adjusted OA prevalence increased in 2005-2018 from 5.37% (95% CI 3.38-7.37%) to 9.19% (5.86-12.52%). For women, age-adjusted OA prevalence among those with the lowest education attainment (less than high school) was lower than in other groups (all P-values ≤ 0.003). We observed significant linear trends among women in all education levels (all P linear trend < 0.04, slope <high school = 1.28, P-value = 0.02; slope high school/GED = 1.14, P-value = 0.004; slope some college = 1.36, P-value = 0.02; slope ≥college = 0.88, P-value = 0.049). Specifically, age-adjusted OA prevalence among women with a high school   The age-adjusted prevalence of self-reported OA by PIR for men and women is shown in Fig. 4. The men with the lowest family income (PIR < 1.3) had a lower age-adjusted OA prevalence among the two PIR groups (both P-values ≤ 0.0003). The age-adjusted prevalence of OA in this group increased from 4.51% (95% CI 3.51-5.51%) to 7.21% (95% CI 5.47-8.95%) in 2005-2012, then remained stable at around 7.5% in the last three survey cycles. Significant linear trends were observed among men with all PIR levels (all P linear trend < 0.03; slope PIR<1.3 = 0.71, P-value = 0.03; slope 1.3≤PIR<3.5 = 0.77, P-value = 0.01; slope PIR≥3.5 = 0.53, P-value = 0.03). Women with the lowest PIR (PIR < 1.3) had a significantly lower age-adjusted prevalence of OA than those people with the highest PIR (P-value = 0.01). During the seven survey cycles, the prevalence fluctuated for women with the lowest family income, but the overall pattern increased. Significant linear trends were observed in all the three PIR groups

Discussion
In this study, with data from a nationally representative sample of US residents in noninstitutionalized populations, we found that women had a higher age-adjusted OA prevalence than men. In addition, significant linear trends and positive slope values of both genders indicate that the OA prevalence increased during 2005-2018. The increasing linear trend in OA prevalence in both genders still remained significant, even after additional adjustments were made in race/ethnicity, educational attainment, and PIR. Moreover, we found statistically significant linear trends in age-adjusted OA prevalence in Non-Hispanic Caucasian and Non-Hispanic African Americans of both genders and in Hispanic women. However, people with lower SES/educational attainment and low PIR reported a lower age-adjusted OA prevalence than people with higher SES. Also, we observed significant linear trends of OA prevalence in most SES subgroups for both genders (all P linear trend ≤ 0.04). The observed trends in OA prevalence among US adults during 2005-2018 were consistent with the study conducted by Dr. Park, which found that the age-adjusted prevalence of OA increased during 1999-2014 7 . Since obesity is a prominent risk factor for OA 19 , the increasing prevalence of this condition among adults in the US might contribute to the rising age-adjusted OA prevalence trend 28 . In our study, the percentage of obesity in both genders increased during 2005-2018, corresponding to the observed increasing OA prevalence trend. Furthermore, our findings regarding a higher prevalence of OA in women than in men also correspond to a www.nature.com/scientificreports/ prior meta-analysis study which found that women are generally at a higher risk of OA than men 29 . Joint space narrowing (JSN) is attributed to the loss of articular cartilage and leads to OA 30 , and women typically have a significantly more progressive decline in joint space than men 31 . Thus, the gender difference of JSN might partially explain the difference in OA prevalence between men and women. In both men and women, the significant linear trend in OA prevalence still exists after multiple adjustments for age, race/ethnicity, educational attainment, PIR, obesity, smoking, and physical activity. Apparently, changes in the distribution of these risk factors cannot fully explain the trend of OA prevalence over the years. In a study by Dr. Dillon et al. in 1991Dillon et al. in -1994, Non-Hispanic African Americans were reported to have a higher prevalence of knee OA than Non-Hispanic Caucasians. African Americans were more likely to have tibiofemoral joint (part of the knee) OA than Caucasians 32 , thus indicating that African Americans had a higher knee OA prevalence than Caucasians. However, in the present study, NHANES did not have information regarding that region of OA. Because our analysis used self-reported OAs, which includes OA in any joints, we could only analyze the prevalence of self-reported OA in any joints. Therefore, our results are different from Dr. Dillon's observations. Caucasians were more likely to have OA on the spine 33 , hand, and other regions than African Americans 34 . In the present study, Caucasians had a higher OA prevalence than African Americans. Notably, we found that age-adjusted OA prevalence was lower among people with disadvantaged SES than people with higher SES. These findings were partially consistent with Dr. Park's findings that OA was more prevalent in older Non-Hispanic Caucasian women with high family income or a college degree 7 . SES is an important determinant of access to healthcare 35 . People with higher SES are more likely to have better insurance coverage for accessing healthcare professionals and will presumably obtain more accurate diagnoses than those with low SES. In the current study, individuals with low SES might lack access to adequate healthcare for OA diagnosis, leading to lower self-reported OA prevalence in that particular group. There are several limitations to this study. First, self-report data of doctor diagnosis were used to define OA in this study because radiographic data were unavailability in NHANES. Also, recall bias possibly impacts the accuracy of prevalence estimates. However, the CDC recommends using self-reported, doctor-diagnosed arthritis as the case definition in estimating the prevalence of arthritis 36 . Studies 28 have proven the validity and reliability of such self-reported data. Second, a small percentage of NHANES participants lack valid information about educational attainment and thus were not eligible for the analysis. In the current study, 0.1% and 9.1% of eligible subjects lacked information about education level and family income, respectively, thus possibly leading to a biased estimate. Third, non-response bias is always a concern in NHANES data, as response rates have declined in federal surveys since 2000 37 . The decline in response rates could have a different impact on OA's estimated prevalence accuracy across the different survey cycles we studied. However, the sample weights of NHANES have accounted for non-response in the analysis. Therefore, these limitations are unlikely to have altered the trends of OA prevalence we observed.

Conclusion
In summary, an increasing trend in OA's age-adjusted prevalence was observed among US men and women during 2005-2018. Non-Hispanic Caucasian and Non-Hispanic African Americans had significantly increasing linear trends in OA prevalence in both genders. People with disadvantaged SES had a lower prevalence of OA. Considering the work limitations and economic burden caused by OA, our findings may be informative in developing related policies to reduce disease development among the population and reduce related risk factors. Our results of OA disparities suggest a need to increase public and health system awareness of OA, especially in Non-Hispanic Caucasian and Non-Hispanic African Americans. Additional research is warranted to further explain the increasing trend in OA prevalence in different races/ethnicities and SES groups in order to more accurately determine the most effective strategies for preventing OA and reducing such glaring disparities. www.nature.com/scientificreports/ Reprints and permissions information is available at www.nature.com/reprints.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.