Understanding atherosclerosis through an osteoarthritis dataset: From knee to vessel

Abstract

Atherosclerotic cardiovascular disease remains a worldwide epidemic and one of the leading causes of death nowadays. Vessel wall imaging can be used to understand the etiology of atherosclerosis but it is rarely done due to the high cost. We recently identified the Osteoarthritis Initiative (OAI), a large prospective cohort study of knee osteoarthritis, which might serve as a valuable source for atherosclerosis research with its serial knee magnetic resonance imaging data. We have found that these images are suitable for vessel wall image analysis of the lower extremity arteries. Here, we will introduce the OAI dataset and explain why it could be used for cardiovascular research purposes. Also, we will briefly comment on peripheral artery atherosclerosis as it is covered in the OAI image dataset and review the use of vessel wall imaging for studying atherosclerosis. We believe data mining of imaging studies, not originally designed on cardiovascular research, can not only maximize the value of the imaging dataset, but also boost our understanding of atherosclerosis.

The Vessels behind Knee: Opportunities from OAI

Cohort studies, as a principal component of study design in modern epidemiology, play a fundamental role in current medical research, including cardiovascular research. Although randomized controlled clinical trials convey the highest quality of evidence, it is not always conductible due to practical or ethical considerations. Additionally, clinical trials are always indicated by observational evidence, which strengthen the need for high quality cohort studies. Running a high-quality cohort study is challenging, even compared with a clinical trial. It usually needs a large sample size to provide adequate statistical power. A prospective design is generally favored over retrospective design to allow for more accurate measurement or determination of the “exposure”. An adequate follow-up time, which is usually very long for chronic diseases, is essential to reach an “outcome”. Moreover, a comprehensive phenotyping of the subjects is desired, despite the additional cost, as it can benefit further data mining as well as control for confounding effects. With these high-demanding requirements for qualified cohort studies, more efficient and effective use of existing data would be an attractive solution. Thus, we propose to use data from the Osteoarthritis Initiative (OAI)¹, supported by the NIH, as it might provide unique insights into atherosclerotic cardiovascular disease (ACVD), beyond its original focus on knee osteoarthritis.

Osteoarthritis and ACVD are two of the most prevalent conditions affecting the US population^{2, 3}. Both conditions are characterized by slow accumulation of pathological change leading, eventually, to adverse health outcomes. They share a number of risk factors including age and obesity. Both conditions develop as a result of a complex and not fully understood interplay between genetics, environment, lifestyle, as well as socioeconomic factors. Successful research addressing these conditions requires prospective studies using large cohorts of at-risk individuals, fully characterized in terms of genetics, demographics, clinical assessment and biomarkers. Interestingly, such studies benefit from intensive use of imaging to characterize the slow accumulation of pathologic change, though it is costly and rarely done^{4, 5}.

The OAI (ClinicalTrials.gov Identifier: NCT00080171) is a multicenter, prospective cohort study of knee osteoarthritis, sponsored by the National Institute of Arthritis and Musculoskeletal and Skin Diseases with additional financial support from several pharmaceutical companies including GSK, Merck, Novartis and Pfizer. It was conceived in the late 1990’s, initiated in 2000, and the live phase ran from 2004 to 2014, with the specific aim to find biochemical, genetic and imaging biomarkers for development and progression of osteoarthritis. The sample size was 4796 and subjects underwent magnetic resonance imaging (MRI) of the knee at 7 timepoints using a rigorously standardized acquisition protocol on four identical Siemens 3T MRI scanners. Participants were recruited into one of three sub-cohorts: the progression cohort included participants with a diagnosis of symptomatic femorotibial osteoarthritis at baseline (N=1390), the incidence cohort included participants at risk of development of symptomatic femorotibial osteoarthritis (N=3284), and the reference cohort included individuals without osteoarthritis and with no relevant risk factors (N=122)⁶. Since its launch, the OAI has been fruitful in osteoarthritis research, with over 500 publications by June 2018, and has become a valuable resource with open access for researchers all around the world. The study design and its dataset lead us to believe that it also offers an exceptional opportunity for cardiovascular research.

OAI recruited a large number of predominantly African American and Caucasian subjects aged 45–79 years, an appropriate age group for studying subclinical and clinical atherosclerotic cardiovascular diseases. The basic demographic characteristics of the OAI dataset are briefly compared to several well-known cardiovascular cohorts in Table 1. The OAI also collected extensive clinical information providing opportunities to address diverse research questions regarding cardiovascular diseases. Cardiovascular risk factor data that have been collected in OAI include age, gender, ethnicity, smoking, body mass index, blood pressure, comorbidity, medications and abdominal circumference. Other data that might be of particular interest to cardiovascular researchers include performance measures (20- and 400-meter timed walk, chair stands timed), quality of life, depression (CES-D), physical activity and extensive evaluation of dietary nutrient intake (Block Brief 2000). More importantly, some of these measures have also been serially assessed during follow-up, enabling investigations into relevant temporal changes. Furthermore, biospecimens including blood (serum, plasma and buffy coat), urine and RNA have been collected at enrollment and consecutively at follow-up visits, serving as a valuable biobank for biomarker studies. Additionally, MRI data of the knee and thigh (both left and right) are available at baseline and follow-up for eligible participants. Because of the high spatial resolution and excellent image quality, it is possible to characterize the popliteal and mid- to distal superficial femoral arteries, and it may be possible to obtain quantitative measurements of the popliteal vessel wall (VW) morphology using methods previously developed and validated for atherosclerotic VW imaging in other vessels⁷.

Table 1.

Comparison of the basic characteristics of OAI to several well-known cohorts

	OAI	MESA	ARIC	CHS
Number	4796	6814	15792	5888
Started	2002	2000	1987	1989
Length of Follow-up	108 months	Ongoing with the 6th Exam (2016–2018)	Ongoing	Ongoing (Most recent available follow-up data through June 30th, 2015)
Age range	45–79	45–84	45–64	≥65
Male	41.5%	47.2%	44.8%	42.4%
Race	79% White or Caucasian; 18% Black or African American	39% White; 22% Hispanic; 28% African American; 12% Chinese-American	73% White and 27% Non-White	16% African American

OAI: Osteoarthritis Initiative; MESA: The Multi-Ethnic Study of Atherosclerosis; ARIC: The Atherosclerosis Risk in Communities Study; CHS: Cardiovascular Health Study.

It should be acknowledged that there are also some limitations of the OAI dataset for cardiovascular research. First, although survival outcome data is available, cause of death was not recorded, and hard cardiovascular outcomes such as myocardial infarction, ischemic stroke, and revascularization were not rigorously collected. The lack of the cardiovascular event data from the study, such as myocardial infarction, stroke, etc., limits its value in establishing the association of potential risk factors with hard cardiovascular outcomes. However, it is possible to use the dataset to study the relationship between baseline atherosclerosis, rate of atherosclerosis progression, other biomarkers, demographic and clinical variables and other health outcomes, such as mental health, as there are data on depression at both baseline and follow-up visits. In principle, comparisons with cardiovascular biomarkers are possible as biospecimens had been collected, although very limited existing measurements are available as no funds were allocated for biospecimen analysis. This could be addressed in the future, cost-effectively, as just an incremental investment on biomarker measurement would be needed for the existing cohort. Second, although a major advantage of MR VW imaging is the identification of plaque components, it is not feasible to analyze the components other than calcification through the 3D double-echo steady state (DESS) sequence used in OAI.

Atherosclerosis in Peripheral Arteries

Atherosclerosis is a systemic pathological process affecting not only the coronary and carotid arteries, but also the peripheral arteries. As with the coronary circulation, atherosclerosis in the lower extremity may result in vascular stenosis with insufficient blood supply to the tissue leading to clinically overt peripheral arterial disease (PAD), presenting ischemic rest pain or intermittent claudication. Symptomatic and / or asymptomatic atherosclerosis is common in the lower limbs and has major prognostic significance⁸. In an international registry of over 50000 patients with established ACVD, one in four patients with coronary artery disease (CAD) also had cerebral vascular disease, PAD, or both⁹. Interestingly, there is a higher prevalence of multi-vascular disease in patients with PAD compared to those with CAD or cerebral vascular disease. In the general population, with aging being one of the most potent risk factors for atherosclerosis¹⁰, the prevalence of PAD was less than 5% among US adults aged 40 years and over, and increased sharply by 3-fold to nearly 15% among those aged over 70 in the National Health and Nutrition Examination Survey (NHANES) study¹¹. Similar trends were also noted in other population-based studies and confirmed by systematic reviews^12–16. Notably, most relevant studies used questionnaires on intermittent claudication or ankle-brachial index (ABI) to determine presence of PAD. However, questionnaires can only identify symptomatic patients and the sensitivity of ABI has been questioned in some cases. In a study using magnetic resonance angiography (MRA), the authors evaluated the relation of ABI to lower extremity artery stenosis, and found that the commonly used cut-off value of 0.9 has a sensitivity of only 15–20% for determining the presence of >=50% stenosis¹⁷. This result is supported by a subsequent systematic review¹⁸. The actual prevalence of peripheral artery atherosclerosis is, therefore, likely higher than reported. In the aforementioned study, the prevalence of high-grade (>=50%) stenosis among a group of subjects aged 70 years is 28% based on MRA¹⁷.

Patients with PAD, either symptomatic or asymptomatic, are at increased risk for future cardiovascular events^19–26. In the REACH (Reduction of Atherothrombosis for Continued Health) registry, those with PAD had the highest cardiovascular mortality, compared to those with established CAD, or cerebral vascular disease or multiple atherothrombotic risk factors⁹. This high cardiovascular mortality in PAD patients may reflect the fact that nearly 60% PAD patients actually had multi-vascular disease in the study. In fact, given the systemic nature of atherosclerosis, presence of atherosclerotic injury at the peripheral vascular bed increases the frequency of disease at other vascular sites, including the coronary artery, the carotid artery and the cerebral arteries^27–29. The prognostic value of PAD may, therefore, be based on its reflection of the underlying systemic atherosclerosis process. This is important since it suggests that studying the peripheral artery can provide reliable information on systemic atherosclerotic burden.

PAD and Osteoarthritis

There is limited information on the association between PAD and osteoarthritis, although there are existing data demonstrating that osteoarthritis is associated with cardiovascular risk. Veronese et al. studied 2158 elderly subjects without cardiovascular disease and found that osteoarthritis at baseline was associated with the risk of incident cardiovascular disease and the association was more prominent in women or when the knee was affected³⁰. PAD was included in the composite cardiovascular endpoint (incident cardiovascular disease) but was not treated as a secondary independent outcome in the study. Gielis et al. found that incident osteoarthritis was independently and positively associated with incident artery calcification in women³¹. The association between osteoarthritis and cardiovascular disease has been confirmed in the OAI dataset³².

For PAD, Park et al. studied the prevalence of lower extremity peripheral vascular disease using ultrasonography in patients undergoing total knee arthroplasty due to osteoarthritis³³. They found that 3.6% of the patients had atherosclerotic changes. This proportion is indeed extremely low especially given that the average age of the study population was 74.1 (range: 65–81) years³³. However, the study excluded those who had vascular-related symptoms (intermittent pain, rest pain, or skin ulcers) and previous history of percutaneous transluminal angioplasty or bypass surgery, and may therefore significantly underestimate the true prevalence of PAD in the general osteoarthritis population.

The explanations for the association between PAD and osteoarthritis remain obscure. Both diseases share a number of common risk factors, including ageing, obesity and chronic inflammation³⁴. These might predispose patients with osteoarthritis to increased risk of PAD, or vice versa. Noticeably, several consequences of osteoarthritis might increase the risk of atherosclerosis, including physical inactivity due to limited mobility, and the use of nonsteroidal anti-inflammatory drugs commonly prescribed for osteoarthritis. The possible role of physical inactivity in osteoarthritis-associated cardiovascular risk has been suggested in two independent observational cohort studies^{35, 36}. A recent randomized controlled trial evaluated the effect of a dosed 12-week walking program on cardiovascular health in patients with severe knee osteoarthritis and found that it helped to achieve targeted blood pressures, as well as to reduce waist circumference³⁷. There is also evidence suggesting that osteoarthritis-associated pain is related to incident CAD³⁸. It is not clear whether this association is a reflection of the link between inflammation and CAD, or if it is due to the effect of stress induced by pain on cardiovascular health.

Vessel Wall MRI for Atherosclerosis Research

Traditionally, angiographic techniques, such as MRA, have been used to measure the degree of lumen stenosis as a proxy for the severity of atherosclerosis. However, atherosclerosis is a disease that resides within the vessel wall rather than in the lumen, with possible expansive, compensatory VW remodeling³⁹. Thus lumen-based disease evaluation will inevitably be insensitive to early vascular insult and miss VW information⁴⁰.

Over the past decade, high spatial resolution MRI has been proven as a reliable method in the study of atherosclerosis, providing information on VW morphology, plaque presence and size, as well as plaque composition^41–50. Particularly, VW MRI has been considered as a cornerstone technique in atherosclerosis imaging research due to its major advantages including: 1. It can measure the thickness of the VW with submillimeter accuracy, enabling the detection of wall thickening before the development of luminal stenosis. This is of particular importance since atherosclerosis in its early phase progresses slowly and silently with wall thickening being the only imaging feature. MRI-derived wall thickness is therefore an attractive biomarker for early disease detection and progression monitoring^{51, 52}. The direct visualization and measurement of VW is also important for detection of positive remodeling, in which the VW expands outwardly and the lumen diameter remains normal⁵³. 2. It can also distinguish plaque components and, therefore, be used to classify the type of the plaque and identify those of high risk⁵⁴. The work of our group has been focused on using VW imaging to determine high-risk plaques and has demonstrated that high-risk plaque components or features (i.e. intraplaque hemorrhage, thinned or ruptured fibrous caps, high lipid content) robustly predict cardiovascular risk compare to plaque burden (number and size) in different disease settings, including PAD^55–59.

MRI has the additional advantage over ultrasound-based VW evaluation (i.e. intima-media thickness, IMT) of being less operator dependent. IMT suffers substantial inter-operator variability, while MRI VW measurements have been reported to have high test-to-test repeatability⁶⁰.

Technically, VW identification on VW MRI relies primarily on fat-suppressed T1-weighted spin echo based black blood techniques. Fat suppression and black blood preparations are used for better delineation of the outer and inner boundaries of the arterial wall. Newer blood suppression methods using motion sensitized dephasing and water excitation, enable the VW to be visualized using gradient echo based DESS MRI in the peripheral arteries^61–63. DESS MRI of the knee was acquired for all subjects in the OAI cohort. While the primary intention of DESS in the OAI study was for quantitative measurements of the knee and articular cartilage, the DESS images also provide an opportunity to measure the popliteal artery VW.

The OAI imaging protocol consists of the following sequences performed on both knees: Coronal and sagittal intermediate weighted 2D TSE and sagittal 3D DESS. Other sequences (coronal T1w 3D FLASH and 2D multi-echo spin echo) were only performed on one knee. The 3D DESS sequence is suitable for VW studies due to: 1) Blood is suppressed due to flow dephasing, 2) fat is suppressed due to water excitation and 3) although acquired sagittally, it can be reformatted to the axial plane with high resolution (acquired at 0.7mm slice thickness and 0.37mm × 0.46mm in-plane). Axial reformats of the sagittally acquired DESS show popliteal VW with good lumen and outer wall boundary contrast (Figure 1). However, it is to be noted that axial reformatting reduces the in-plane resolution to 0.7mm × 0.37mm while improving the slice resolution to 0.46mm. While artery segmentation is easier using axial reformats, there may be potential advantages in segmenting directly using the sagittal images due to higher in-plane resolution. The entire extent of the femoropopliteal artery in the large sagittal field-of-view (140mm), centered on the knee, can be targeted for VW measurements. The popliteal artery is conventionally divided into three sections P1, P2 and P3⁶⁴, of which the whole of P2 and parts of P1 and P3 including the bifurcation are covered by this field-of-view. Detailed imaging parameters for 3D DESS and other OAI protocol sequences can be found in a previous report⁶⁵.

Figure 1:

Axial cross section through the popliteal artery showing wall thickening in one OAI subject. Right panel shows enlarged popliteal artery with and without contours (lumen contour in red, outer wall contour in blue).

As aforementioned, our initial inspection of the axial 3D DESS images suggests that it is feasible to obtain quantitative measurements of the popliteal VW using established methods. Using expert reviewer assisted segmentation of the lumen and outer wall in the CASCADE plaque analysis software (University of Washington, Seattle)⁷, measurements such as lumen area, percent stenosis, wall area and wall thickness can be obtained in addition to atherosclerotic plaque composition if present. It has to be noted that while calcification can be identified using 3D DESS, identification of other plaque components using 3D DESS have not yet been established. Therefore, it is unfortunately not feasible to depict detailed plaque component in the OAI dataset.

The image quality for applying established VW analysis methods also appears to be excellent. In a randomly selected set of 38 knees, we rated the image quality (scale of 1 to 3 with 3 being the best), presence of flow artifacts and presence of bifurcation/trifurcation of the popliteal artery. The mean image quality was 2.29±0.47. There were minor flow artifacts in 21/38 knees but did not affect the visualization of lumen boundaries.

A particular advantage of the OAI imaging data lies in its serial imaging, with annual imaging for 4 years after enrollment and biannual imaging from the 4^th to the 8^th year. A total of 7 time-point image sets, including baseline, were collected during that time-frame with identical imaging protocol, coverage and quality control. This compares favorably with most cardiovascular-specific cohorts, in which cardiovascular imaging was repeated less frequently or was just collected at one-time point. Serial measurement is particularly important in view of the fact that atherosclerosis is an evolving pathologic process. While atherosclerotic plaques remain asymptomatic for long periods, they are not static. Instead, lesions continually undergo a dynamic process involving both tissue injury and repair. Consequently, progression, as well as regression, of plaques has been observed in the patients^66–69. In light of this, serial imaging over time is preferred due to the following advantages: 1) it aids in elucidating the temporal relationships of changes in risk factors and changes in atherosclerotic lesions, and 2) it allows for the evaluation of the potential effects of particular risk factors or interventions over time (e.g. yearly in OAI).

The highly detailed sequential image dataset in OAI will permit a comprehensive and highly quantitative assessment of the temporal evolution of atherosclerosis in vivo, including response to clinical or lifestyle changes. Given the high cost of MR imaging studies, it is unlikely to be superseded in the foreseeable future.

OAI for Cardiovascular Research

By the end of 2018, there have been over 500 publications based on OAI, with the MR imaging data contributing to a significant portion^{70, 71}. Naturally most of those publications were focused directly on osteoarthritis, with little attention to cardiovascular risk or disease.

Data from the OAI has provided invaluable insight into the relationship between risk factors and specific osteoarthritis outcomes, as well as relationships between risk factors and morbidity and mortality outcomes in general. The length of the study and frequency of follow-up allowed some “natural experiments”, where the benefits of, for example, weight loss or changing levels of activity during the study, could be determined. Factors found to increase the risk of osteoarthritis incidence or progression included abnormal glycated serum protein, physical inactivity, slow gait, low dietary fibre intake, soft drink consumption, high saturated fat intake, high total fat intake, vitamin D deficiency, high baseline body mass index, and analgesic use^72–83. Factors found to decrease the risk of osteoarthritis incidence or progression included weight loss during the study, high intake of mono- and poly-unsaturated fatty acids or adherence to a Mediterranean diet, and mitochondrial haplotypes T and J^{79, 84–88}. Factors not found to increase the risk of osteoarthritis incidence or progression include depressive symptoms at baseline, a history of running, statin use, and smoking history^89–92. More generally, the health benefit was observed to inactive older adults of increasing physical activity , while health risks were associated with sedentary behaviour, slow gait, obesity or central obesity^93–98. Elevated mortality was associated with consumption of fried potato products⁹⁹. Since many of these risk factors are also of interest in cardiovascular research, it may be illuminating to determine their effects on atherosclerosis initiation and progression, as quantified by vessel wall MRI.

One study specifically utilized the dataset for mining cardiovascular outcome (incident cardiovascular disease) and looked into the relationship between baseline osteoarthritis and future risk of cardiovascular disease³². The authors concluded that hand osteoarthritis predicts the risk of developing cardiovascular disease in women, but not in men. However, the lack of endpoint ascertainment for the self-reported onset of cardiovascular diseases and the large proportion of participants loss to follow-up posed significant limitations to such study. Another study group performed a cross-sectional analysis of the association between sedentary behavior and blood pressure control using the 48-month visit data, in which accelerometers were used to measure physical activity, and demonstrated a graded association between sedentary time and elevated blood pressure among participants who were not taking antihypertensive medication¹⁰⁰.

Conclusions and Perspectives

Atherosclerosis remains the major leading public health burden globally. A thorough understanding of the disease’s natural history in vivo is essential for risk stratification and identification of novel targets for prevention and treatment. Serial atherosclerosis imaging study provides a critical tool for early detection and monitoring of disease development and progression in such studies. MR-based VW imaging is an ideal research tool for this purpose. The OAI is a large longitudinal study of knee osteoarthritis with annual MR imaging data of the knee and thigh for up to 8 years. We recently identified that these MR images could also be used for popliteal and superficial femoral arterial wall analysis. By leveraging its large sample size, available information on cardiovascular-related risk factors, and high-fidelity MR VW imaging data, the OAI could provide an unexpected and valuable opportunity for studying the natural history of atherosclerosis.

Highlights.

• In this review, we propose to use a large, prospective, serial, bilateral knee MRI originally obtained to study osteoarthritis to now characterize lower extremity atherosclerosis by measuring the vessel wall of the femoral-popliteal artery.

• We demonstrate the feasibility of such a study and discuss why this repurposing can enhance our understanding of atherosclerosis.

• The leverage of the serial MR imaging of this osteoarthritis cohort will provide a unique opportunity to study the temporal resolution of atherosclerotic lesions in a large population.

Abbreviations

ABI

ankle-brachial index

ACVD

atherosclerotic cardiovascular disease

CAD

coronary artery disease

DESS

double-echo steady state

IMT

intima-media thickness

MRA

magnetic resonance angiography

MRI

magnetic resonance imaging

OAI

Osteoarthritis Initiative

PAD

peripheral arterial disease

VW

vessel wall