Background
Osteoarthritis (OA) is the most common form of arthritis and has been reported to affect around 12% of the adult population in both the United States [
1] and the United Kingdom [
2]. This high prevalence is associated with significant economic [
3] and social [
4] burden, which is expected to increase with an ageing population and increasing obesity [
5].
The definition of OA for epidemiological studies and estimates of prevalence have been areas of established interest. Previous community studies have used radiographs as the main imaging modality to define OA at the hand [
6‐
10], knee [
11‐
15] and hip [
10,
15‐
17], despite recognised limitations [
18]. Radiographs have shown poor correlation with clinical symptoms [
19,
20] and might not be adequately sensitive as an outcome measure in clinical trials [
21]. Ultrasound is an imaging modality that has been gaining increasing popularity in the assessment of OA [
22‐
24]. The advantages of ultrasound are well described; most importantly it is relatively low cost, non-invasive and lacks exposure to ionising radiation. Ultrasound can identify soft tissue structures such as the synovium; important because synovial inflammation is known to be a predictor of progression of knee OA [
25]. The multi-planar ability of ultrasound adds dimensions to imaging that are not possible with radiographs and may lead to easier recognition of osteophytes than when compared to radiographs [
26]. The use of ultrasound might also help to explain the reported discordance between clinical symptoms and imaging in OA. We have previously demonstrated the reliability and validity of ultrasound in detecting features of knee OA in community participants [
27]. Thus, ultrasound imaging has the potential to become an extremely useful tool in studying features of OA in the community.
The hypothesis of this study was that the use of a more sensitive imaging modality (ultrasound) would lead to a different (most likely higher) prevalence estimate for OA at the dominant hand, knees and hips when compared to previous radiographic studies. Osteophytes were imaged at the hand, knees and hips; femoral head abnormality at the hips and synovial effusion at the knees, since these are ultrasound features of OA that have a definite cut-off to define presence or absence. The Newcastle Thousand Families birth cohort presented an ideal setting of unselected individuals from the community to perform this study.
Discussion
The estimated prevalence of features of OA in this cohort was 78% for the dominant hand, 30% for knee and 41% for hip. We have, demonstrated a higher prevalence of OA using ultrasound compared to previous studies using radiographs, especially in the hand and the hip joints.
For knee OA, the prevalence estimates of this study are not dissimilar to that found in the 55–64 year age group of the Johnston County Osteoarthritis project [
12], which had a high proportion of overweight (74% had a BMI ≥ 25 kg.m
-2; compared to 61% in our cohort). The prevalence estimate of 33% radiographic knee OA in the Framingham study [
11] is also similar to our study. However, the Framingham cohort had participants with a mean age of 73 years; which is ten years older than the Newcastle Thousand Families cohort currently. It is possible that the healthier lifestyles that may have been prevalent at the time of the Framingham assessment (~30 years ago) would have brought the prevalence estimate of OA down in that cohort. A study of 1729 community dwelling individuals from Nottingham, UK, with a mean age of 64 years (mean BMI 26.8) estimated the prevalence of symptomatic radiographic knee OA to be much lower at 11%, despite the cohort including participants from a previous knee study on pain [
10]. The differences in the above estimates could also be explained by the fact that the above three studies used different criteria to define radiographic knee OA. The Nottingham study required the presence of knee pain with a grade 2 osteophyte, while the Framingham and Johnston County studies required a Kellgren-Lawrence (K-L) grade of ≥ 2. The lower prevalence of OA in the Nottingham cohort could therefore be explained by the need for symptoms (knee pain) in addition to radiographic features of OA.
A further report from The Johnston County Osteoarthritis Project study of 2637 community dwelling men and women aged 45 years and older showed an overall prevalence of radiographic hip OA of 27.6% [
16]. The prevalence increased with age; the 55–64 year group had a prevalence of 23% and this increased to 31% in the 65–74 year age group. These estimates were higher than the previously published NHANES 1 study [
37]. The authors of the Johnston County study suggested that this might have been due to a higher prevalence among rural (vs urban) participants in Johnston County. Other reasons include the possibility of changes in hip OA risk factor profiles over time as well as variation in radiographic techniques and interpretation. While the definition of radiographic hip OA in the Johnston County study and NHANES 1 were the same (K-L grade of ≥ 2), it is still feasible that the images from the NHANES 1 study were under-read. One of the reasons for the high prevalence of hip OA noted in the Newcastle Thousand Families cohort could have been because the ultrasound definition for presence of an osteophyte was more sensitive than the radiographic K-L score of ≥ 2 (which requires a “definite” osteophyte, as opposed to the “doubtful” osteophyte seen with a K-L score of 1).
1467 men and 1519 women of the Medical Research Council’s (MRC) National Survey of Health and Development had a clinical hand examination performed at age 53 [
6]. The prevalence of OA (defined by examination, but not by imaging) was 21%, 12% and 8% in any of the DIP, PIP and CMC joints, respectively, among women. In men, the corresponding values were 14%, 8% and 4%. However, the participants were 10 years younger than the Newcastle Thousand Families cohort were when examined and were therefore likely to have lower prevalence estimates. Another reason for the low estimates in the MRC survey is that ultrasound is likely to be more sensitive than clinical examination in the detection of osteophytes. A cohort of 489 participants from a family study of nodal OA in Nottingham in 2004, had a prevalence estimate for radiographic OA of 46% in the right index finger and 73.3% for overall hand OA [
10], results quite similar to our study. The mean age of the Nottingham cohort was 65.7 years with a high proportion of women (82.6%). However, this was not a random population sample as the Nottingham participants were selected for their higher risk of prevalent hand OA. Factors known to influence the prevalence of hand OA such as occupation [
38], grip strength [
39] and other systemic factors such as use of oestrogen [
40] and obesity [
41] might be responsible for some of the differences observed.
In contrast to the Nottingham radiographic cohort study [
10], we did not find any significant differences between the right and left knee for osteophyte detection using ultrasound. Similar to their findings though, there was no significant difference between the right and left hip osteophyte prevalence. Our findings are, however, in line with the Zoetermeer Survey that showed no evidence of any right-left differences of radiographic OA at the knee and hip joints [
14]. This suggests that bio-mechanical stresses act equally on both sides of the lower limb joints. Equally, it might also suggest that genetic and systemic factors such as oestrogen [
40] and obesity [
41] might play a larger role in the pathophysiology of OA at the knee and hip, than bio-mechanical stress. We report that the prevalence of isolated knee and/or hip OA was extremely low in this cohort (8%); while isolated hand OA was particularly high (31%). This high prevalence of hand OA suggests that ultrasound defined hand osteophytes may be a predictor of a more generalised form of OA affecting the knee and the hip. Cooper et al. [
42] demonstrated that there is no single threshold number of joint sites that could be used to define generalised OA. However, for the purpose of this study, we used a definition that would include two out of three joint sites (hand, knee and hip). However, a longitudinal follow-up of these participants will be required to consolidate this hypothesis, since it is possible that biases such as a survival bias might have influenced the cross sectional results noted above.
We did not find a higher prevalence of knee OA in women, as has been noted in previous studies, such as the Framingham study [
43]. However, the Framingham study showed an increasing prevalence of OA with age, particularly in women. The Framingham participants were aged 63 years and older (and hence older than the Newcastle Thousand Families Cohort) and the higher prevalence of knee OA in women in Framingham was seen mostly in the older age groups. The prevalence of knee OA in the 63–69 year age group was higher in men (30%) when compared to women (25%) but in the ≥80 year age group, there was a higher prevalence seen in women (53%) when compared to men (33%). This might explain why the prevalence of knee OA was not higher in females in the Newcastle Thousand Families Study, as the participants were aged only 63 years and it is likely that at this (relatively young) age there is no significant difference in knee OA prevalence between the sexes. Indeed, a meta-analysis study of sex differences in knee OA demonstrated that among those aged ≥55 years of age, there was a significantly lower pooled risk of prevalent radiographic knee OA, but this significance was lost when looking at those <55 years of age [
44].
The prevalence of knee effusions was remarkably high in this study (just under a quarter of knees had ≥4 mm knee effusion on ultrasound), considering the subjects were a population sample, and not selected for symptoms. The high prevalence could potentially be explained by defining effusion with a low cut-off (i.e. 4 mm). However, this definition has been validated against symptoms in a multi-centre European study [
34]. It was also interesting to note that males had a trend towards a higher prevalence of knee effusion than females, although this did not quite reach statistical significance (p = 0.1). Since there is increasing evidence that inflammation predicts knee OA progression [
25,
45], this would suggest that males in this cohort might be at a higher risk of rapidly progressive OA. Further longitudinal follow up of the cohort should help to address the issue of predictive validity of knee effusion identified by ultrasound.
A strength of our study is that it uses a well characterised population-based cohort to estimate the prevalence of features of OA at different joint sites using a sensitive imaging modality, namely ultrasound. This is the first study to estimate prevalence of certain ultrasound features of OA in the community. Although there were a few variations in inter-rater agreement across the joints, the level of agreement found on inter-rater reliability of ultrasound images in this study helps to reduce some of the concern about the subjective nature of ultrasound scoring of images.
There were a few limitations to this study. We mainly imaged the bony parameters in the joints and did not include prevalence estimates for cartilage thickness. This is because of the lack of a clear cut-off in the knee to define “cartilage thinning” as well as the limited ability of ultrasound to identify cartilage morphology in the hand joints due to technical feasibility. However, this would suggest that the prevalence estimates of knee OA in our study are likely to be under estimates. The study populations, definitions and image acquisitions in previous radiographic studies of OA have varied considerably [
46] and hence any direct comparisons with this study are difficult to make. Another limitation is that radiographic osteophytes were not reported and hence comparison of radiographs with ultrasound imaging was not done. Power doppler synovitis was not assessed in this study, as the knee and hip joints are considered too deep for accurate assessment of power doppler signal and power doppler in hand OA has not yet been shown to demonstrate adequate construct validity. Kappa statistics for inter-rater reliability of ultrasound imaging were only calculated on a small proportion of the participants. This might explain the uncertain estimates and large confidence intervals noted at a few of the joint sites. Furthermore, the inter rater reliability at the hip was moderate in this study, rather than the substantial or excellent reliability at other sites. Further work to standardise the acquisition techniques and reading of hip ultrasound images for features of OA will help to decrease imprecision of prevalence estimates due to the imaging modality itself. Also, this study was performed among members of a birth cohort born in the city of Newcastle upon Tyne, UK, which might reduce the external validity of the findings. Another drawback is that inferences can only be made for this particular age group of subjects (aged 61–63 years). However, this is an age group that has particular public health importance, as the prevalence of OA in this age group is quite high and effective risk factor modification (e.g. diet and exercise) is still likely to be feasible.
Competing interests
The authors declared that they have no competing interests.
Authors’ contributions
MSP, FB and RMF conceived and designed the study, and obtained the funding. AMA carried out the ultrasound assessments and MSP was responsible for directing both the overall clinical assessment and the data collected by questionnaire (the aspects included in this paper were designed by MSP, AMA, FB and RMF). AMA and KDM did the statistical analysis, supervised by MSP. AMA and MSP drafted the paper with critical contributions from all other authors. All authors read and approved the final manuscript.