Abstract
Objective. The purpose of this study is to examine the course of hand osteoarthritis (HOA) and its relationship with cardiovascular disease (CVD) and diabetes (DM).
Methods. Data were collected at 3 timepoints from 845 Johnston County Osteoarthritis Project participants (two-thirds women, one-third African Americans, mean age 60 yrs) with and without HOA, CVD, or DM. A diagnosis of radiographic HOA (rHOA) required a Kellgren-Lawrence severity grade of ≥ 2 in at least 3 joints in each hand. A 4-state progressive model included transitions based on rHOA and pain or function as defined using the Australian/Canadian HOA Index (AUSCAN). Markov multistate models estimated HR (aHR) and 95% CI for associations between DM or CVD and specific state transitions, adjusting for baseline and time-varying covariates.
Results. Participants with DM (vs those without DM) were more likely to experience worsening pain with rHOA. Individuals who had or developed CVD (vs those who did not) were significantly less likely to experience symptomatic improvement, regardless of rHOA status. Those with DM or CVD (vs those without these comorbidities) were less likely to experience improvement in function, although this was statistically significant only for those with DM and no rHOA.
Conclusion. Overall, having or developing DM and/or CVD reduced the likelihood of symptomatic and functional improvement over time, suggesting an effect of comorbid CVD and DM on the clinical and radiographic course of HOA. Additional studies are needed to confirm these findings.
Hand osteoarthritis (HOA) is a highly prevalent condition, can affect 1 or more joints of the hand, and frequently leads to clinical symptoms and physical limitation1. Studies of the prevalence of HOA show great variation. For example, in 2011, the Framingham Osteoarthritis Study2 reported an HOA prevalence of 13% in men and 26% in women > 70 years of age, while findings from a study in China showed much lower prevalence values (3.0% of men and 5.8% of women)3. More recent estimates from the Johnston County Osteoarthritis Project (JoCo OA) set the lifetime risk of developing symptomatic HOA at 40%, further demonstrating the public health significance of this condition4.
Known risk factors for HOA include advancing age and female sex. Obesity has been implicated with OA pathology, especially in the knee and hip joints, where it is known to manifest, at least in part, through a body mass index (BMI)-driven mechanical stress on the joint5. Obesity is associated with HOA as well, although less consistently6. Because hand joints are not classified as weight-bearing in character, a more systemic, inflammatory mechanism for HOA has been proposed7. The relationship between metabolic syndrome (MetS), a condition including hyperglycemia, insulin resistance, obesity, and dyslipidemia, and HOA has been the subject of several studies; some have demonstrated a lack of association between MetS and incident HOA or symptomatic HOA8,9, while others demonstrate positive relationships between MetS components and HOA10–17. Additionally, clinical symptoms, physical function, and radiographic changes associated with HOA have been investigated, finding that limitation in physical function correlates with increasing hand pain18,19.
The association between cardiovascular disease (CVD) and/or diabetes mellitus (DM), and HOA remains unclear. HOA has been positively associated with higher prevalence of atherosclerosis and endothelial dysfunction20,21,22,23,24, coronary heart disease9, CV events14,25, and overall CV-related mortality26. Regarding DM, some studies have shown no relationship between HOA and type 1 or 2 DM9,14,27,28,29 or impaired glucose metabolism30, but 1 cross-sectional study31 found patients with DM had more than double the odds of hand or knee OA, while another study showed an increased prevalence of hand enthesophytes, or bony spurs associated with HOA, in individuals with DM32. To date, research in this area has been limited by a lack of standardized methods across studies for assessing DM or CVD presence and measuring and quantifying HOA presence and severity, as well as differences in study populations, sources of participants, duration of follow-up, and overall study design.
Our study objective was to further assess associations between radiographic changes, pain, and function in HOA and self-reported DM and CVD in a large, longitudinal community-based cohort using standardized assessments and an advanced statistical approach, in a hypothesis-generating manner.
MATERIALS AND METHODS
Study participants
We used data from the JoCo OA, a large, prospective community-based cohort, well-characterized for a variety of OA-related risk factors and outcomes as well as general health and comorbid conditions. The details of this cohort are described elsewhere33. In brief, an original (baseline data collection between 1991 and 1997) cohort and enrichment (baseline data collection between 2003 and 2004) cohort were enrolled and followed about every 6 years. Follow-up data were collected between 1999 and 2003 for the original, and 2006–2011 and 2013–2015 for both the original and enrichment cohorts. At baseline, participants were noninstitutionalized civilian African American (oversampled to equal roughly one-third of the cohort) or white men or women, adults ≥ 45 years of age, residents of 1 of 6 chosen townships of Johnston County, North Carolina, as detailed elsewhere4,33. HOA symptoms and radiographic measurements were not collected at baseline for the JoCo OA; thus for our study, we analyzed data from the 3 subsequent study visits (the baseline visit was either the 1999–2003 follow-up or the 2003–2004 cohort enrichment). The study has been continuously approved by the University of North Carolina Institutional Review Board (IRB# 920583).
Covariates
The JoCo OA protocol consisted of both home interviews and clinic visits. Static covariates included age, sex, race, and education level (categorized as < 12 or ≥ 12 yrs of education). Covariates treated as time-varying in our analysis, collected at all 3 timepoints, were BMI, symptomatic knee OA (sxKOA) status, nonsteroidal antiinflammatory drug (NSAID) use, and self-reported DM and CVD status. BMI was derived from height and weight measurements at each clinic visit, with obesity defined as a BMI of ≥ 30 kg/m2. We determined NSAID use (any type) from review of medication lists. SxKOA was defined as (1) a Kellgren-Lawrence score ≥ 2 in at least 1 knee [by a single, experienced musculoskeletal radiologist (JBR) with high reliability33] and (2) in the same knee, an answer of yes to the question, “On most days of any one month in the last 12 months, did you have pain, aching or stiffness in your right/left knee?” For defining CVD and DM, participants were read the following statement: “Please tell me which of the following conditions or illnesses a doctor, nurse, or health professional has told you that you have now or have ever had.” Self-reported DM status was then elicited through a yes/no answer to “diabetes or high blood sugar.” Across data collection cycles, CVD status was assessed with increasing inclusiveness and specificity (baseline: heart attack, other heart problems, cerebrovascular accident; first follow-up: added angina, congestive heart failure; second follow-up: added peripheral vascular disease). We assessed each patient’s number of clinical comorbidities, for a maximum of 3 (DM, CVD, and obesity), at every timepoint.
Hand symptoms and rHOA assessment
The same radiologist (JBR) assessed each of the 30 joints of both hands on posteroanterior radiographs for signs of rHOA by using a standard OA atlas34. Hand radiographs from all 3 timepoints were assessed simultaneously in known time order. Reliability of rHOA assessment by the radiologist was previously demonstrated to be good to excellent4. Participants were classified as having rHOA if they met the Genetics of Generalized Osteoarthritis (GOGO) study criteria35: Kellgren-Lawrence severity grade of 2 or more in at least 3 joints with 2 of these 3 in the same joint group [distal interphalangeal joint (DIP), proximal interphalangeal joint (PIP), and/or carpometacarpal joint] across both hands, involving at least 1 DIP, and with no more than 3 swollen metacarpophalangeal joints36. For classification purposes, the thumb interphalangeal joint was categorized as a PIP joint4.
Pain and function were assessed using the Australian Canadian Osteoarthritis Hand Index (AUSCAN), a validated self-report measure of hand pain, stiffness, and function with a question response range of 0–4 (none to extreme)37,38. The pain (5-item) and function (9-item) subscales were considered individually for analysis.
Statistical analysis
Descriptive statistics were calculated as frequencies and percentages for categorical variables, and as means and SD for continuous variables. The AUSCAN pain subscore values were dichotomized using upper and lower quartile cutoffs (at the 75th percentile, or a numeric score of 6), as noted in previous literature18 and supported by close correspondence between these quartile values and a calculated pain subscore based on all “mild” responses. An AUSCAN pain subscore ≤ 6 was defined as “low pain” and an AUSCAN pain subscore > 6 was defined as “high pain.” This same method was used in assessing functional limitation and establishing a cutoff value for the 9-question AUSCAN physical function subscale (at the 75th percentile, or a score of 9). An AUSCAN function subscore ≤ 9 was defined as “better function” and an AUSCAN function subscore > 9 was defined as “worse function.”
At the participant level, and if present in at least 1 hand, the following 4 GOGO rHOA and AUSCAN pain states were defined: (A) No rHOA with low pain; (B) rHOA with low pain; (C) no rHOA and high pain; and (D) rHOA and high pain (Figure 1). Similar states were defined using AUSCAN functional cutoffs (Figure 2). Types of modeled transitions (i.e., transitions among consecutive visit pairs) can be summarized as follows: incident rHOA (state A to state B, and state C to state D), worsening pain/function (state A to state C, and state B to state D), and improvement in pain/function (state C to state A, and state D to state B). Transitions from a state of rHOA to no rHOA (state B to state A, and state D to state C) were not modeled (not a common or clinically relevant change).
To assess the independent association of comorbid conditions with changes in HOA and symptom characteristics, a time-to-event approach was used to allow for (1) study of hand event outcomes with nonexact times of transition (managed as interval-censored) and time-dependent main effects, and (2) multiple hand event states of interest. We used Markov multistate survival models (MSM) for interval-censored outcomes using parametric, piecewise constant exponential distributions to define a 4-state progressive model and transitions of interest (above and Figure 1). MSM use the theory of stochastic processes (a set of random variables representing the evolution of a process over time) to assess transitions between states in continuous time. Under the Markov assumption, future transitions were assumed to depend only on the current state, independent of time. The MSM were used to estimate HR and 95% CI for each comorbid condition and the 6 transitions simultaneously. All models were person-based and adjusted for aforementioned static covariates (e.g., sex), or time-dependent covariates, which were allowed to change at observed times (e.g., BMI). Each comorbid condition was examined individually in an overall model. In a separate model, a count variable of comorbid conditions of DM, CVD, and obesity, ranging from 0 to 3, was assessed for its association with HOA states. All MSM analysis was conducted using R software through the msm package (R Foundation for Statistical Computing)39.
RESULTS
A total of 845 participants from the JoCo OA longitudinal cohort had all baseline covariates and comprised the complete case analysis sample (Figure 3). The 2 study follow-ups were at a mean ± SD of 6.4 ± 0.8 and 12.0 ± 1.2 years from baseline, respectively. At baseline, the mean age was 60 ± 7.4 years, with about one-quarter of individuals age 65 years or older. Two-thirds were women and nearly one-third were African American (Table 1A). The average BMI of the study population at baseline (30.9 ± 6.5 kg/m2) was at the lower end of the obese classification40, and remained relatively consistent over the study intervals (Table 1A). One in 5 participants self-reported CVD at baseline, increasing to over 40% at the second follow-up. Similarly, about 10% of individuals had DM at baseline, with this percentage increasing to 3-fold by the second follow-up. About 15% of the study population had sxKOA at baseline, increasing to 25% by the second follow-up. Finally, NSAID use (any, regardless of type, purpose, or duration) was reported by almost half of study participants at baseline with over 2 of 3 individuals using NSAID by the final data collection cycle. For the outcomes, 24% of the participants had high hand pain and 8% had rHOA at baseline. At baseline, median (interquartile range) was 0 (0–6) for AUSCAN pain and 1 (0–9) for AUSCAN function.
Transitions defined by rHOA and AUSCAN Pain subscores
The majority of individuals remained in state A without rHOA and low pain (Table 1B). Because an individual could contribute or make multiple transitions during follow-up, the number of transitions is not equivalent to the number of participants (Figure 1).
For incident rHOA, there were 77 transitions to incident rHOA while maintaining low pain (state A to state B) and 22 transitions to incident rHOA while maintaining high pain (state C to state D). No significant associations were seen between DM or CVD and the transition to incident rHOA, although some comparisons were limited by small numbers (Table 2).
For worsening pain (from low to high pain), there were 123 transitions in those without rHOA (state A to state C) and 36 in those with rHOA (state B to state D). No significant associations were seen between CVD and worsening of AUSCAN pain subscore (Table 2). However, there was a statistically significant association between having or developing DM and worsening pain in individuals with rHOA (state B to state D; Table 2).
For improvement in pain (from high to low pain), there were 126 transitions in those without rHOA (state C to state A) and 20 in those with rHOA (state D to state B; Figure 1). As shown in Table 2, individuals who had or developed CVD during the course of the study, versus those who did not, had a significantly lower hazard of improvement in pain regardless of rHOA status (state C to state A: aHR 0.58, 95% CI 0.36–0.92; state D to state B: aHR 0.13, 95% CI 0.02–0.77).
Transitions defined by rHOA and AUSCAN function subscores
For incident rHOA, there were 73 transitions to incident rHOA while maintaining a better functional status (state A to state B) and 24 transitions to incident rHOA while maintaining a worse functional status (state C to state D; Figure 2). No significant associations were seen between DM or CVD and the transition to incident rHOA (Table 3).
For worsening function (from better to worse function), there were 141 transitions in those without rHOA (state A to state C) and 37 transitions in those with rHOA (state B to state D). No significant associations were seen between DM or CVD and worsening function, regardless of rHOA status (Table 3), although there was a borderline association between having or developing CVD and risk of worsening function in the setting of rHOA (state B to state D: aHR 3.55, 95% CI 0.88–14.3).
For improvement of function (from worse to better function), there were 119 transitions in those without rHOA (state C to state A) and 13 transitions in those with rHOA (state D to state B; Figure 2). Compared to individuals without DM, those with DM and no rHOA had a 42% lower rate of improvement of hand function (state C to state A; Table 3). Those with CVD and no rHOA, compared to those without CVD, also showed a borderline significant reduced hazard for improvement in AUSCAN function subscore (state C to state A; Table 3).
When a count of comorbidities (obesity, DM, and CVD) was used, those with 1 or 2 comorbidities compared to those with none had about half the hazard of improvement in pain in the absence of rHOA [state C to state A, aHR for 1 (95% CI): 0.53 (0.32–0.89); aHR for 2: 0.57 (0.31–1.04)]. Those with all 3 comorbidities had over 60% reduced hazard for improvement in pain (aHR 0.34, 95% CI 0.13–0.88). Those with 1 or 2 comorbidities, compared with none, also had a reduced hazard for improvement in function without rHOA [state C to state A, aHR for 1 (95% CI): 0.60 (0.36–1.02); aHR for 2: 0.55 (0.30–1.01)]. Additionally, those with 1–2 comorbidites had about 4 times the hazard of developing worsening function in the setting of rHOA [state B to state D, aHR for 1: 4.05 (1.52–10.84); aHR for 2: 3.98 (1.07–14.85)]. This was not seen in the AUSCAN function model for those with 3 comorbidities.
DISCUSSION
In our study, we examined the effects of DM and CVD on the radiographic, symptomatic, and functional transitions of HOA. We found that comorbid DM increased the odds of worsening AUSCAN pain among those with rHOA, while comorbid CVD significantly reduced the chance of AUSCAN pain score improvement in those with and without rHOA. Improvement in hand function was less likely in those with DM or CVD, compared to those without these comorbidities. Finally, the concurrence of multiple comorbidities (obesity, DM, and CVD) was associated with (1) less likely improvement in pain or functional status in those with rHOA, and (2) more likely worsening function in the presence of rHOA.
The imperfect concordance between hand pain/disability and the presence of rHOA19 suggests a need for further investigation into other determinants (e.g., CVD or DM) of the HOA disease process41. Pain can influence the course of OA, acting as a mediator in the loss of hand function18. Patients reporting high levels of pain at baseline in 1 study exhibited greater functional limitation and worse pain outcomes in the long term, with only a quarter of individuals experiencing an improvement in HOA symptoms over 6 years19. Another study demonstrated that progression of pain and functional limitation can occur in many HOA patients over just 2 years42.
In our study, HOA patients with comorbid DM experienced an increased hazard for worsening pain in the presence of radiographic changes. There is a lack of consensus regarding the association between DM and HOA, due to differences in definitions and cohorts. An increased prevalence of hand pain, disability, and stiffness not explained by radiographic changes or peripheral neuropathy in a group of patients with type 1 DM was seen in the Dialong hand study27. DM was significantly associated with an increased prevalence of symptomatic HOA in a Chinese Han population20. On the other hand, findings from the Netherlands Epidemiology of Obesity study demonstrated no connection between DM and AUSCAN pain in cross-sectional analysis of all participants30. A systematic review43 on DM as a risk factor for HOA revealed no studies that indicated DM as an independent risk factor for HOA. Study designs differ; the NEO study was cross-sectional6. Courties, et al9 incorporated a 3-year longitudinal follow-up of only rHOA patients, compared to our 6- and 12-year follow-up of individuals including those without rHOA. Several studies have demonstrated a relationship between DM and erosive HOA, a more progressive and severe form of the disease27,30,44.
Research into the co-occurrence of CVD and HOA in patients has also given inconsistent results. CVD-related comorbidities have been closely linked to symptomatic HOA14 and even proposed as independent risk factors for HOA development45; symptomatic HOA was also associated with an increased prevalence of coronary heart disease in the Framingham Heart Study25. Other studies have found no association13,46,47. Atherosclerosis and resultant coronary heart disease has been associated with rHOA severity22, contrary to our study findings; however, previous associations with increased AUSCAN pain score and deterioration of Functional Index of Hand Osteoarthritis score9 concur with our results. Several pathophysiological mechanisms have been proposed. It is known, for example, that in patients with OA, inflammatory mediators are released from the joint into systemic circulation, thereby rendering a low-grade inflammation thought to induce atherosclerosis48. Given the many connections and proposed mechanisms, and the complex relationship between HOA and CVD, this is an area in need of further research.
Because little is known about the clinical course and progression of pain and functional limitation in patients with HOA42, the longitudinal design of our study, and the relatively large sample size, stand out when compared to many of the prior longitudinal and cross-sectional investigations. The JoCo OA study design also allowed for analysis of time-varying covariates at each timepoint, thus providing a more complete picture of the interplay between comorbidities and clinical and radiological course of HOA. However, this work also has limitations. Despite our overall large sample, small numbers were seen for several of the transition states, complicating analysis and interpretation of the findings and leading to large CI; therefore, the study is hypothesis-generating. The self-reported aspect of the comorbid conditions and the medication lists (generic or brand name) also could be a limitation, although likely less so for DM (where positive and negative predictive values compared to medical records in prior studies have been > 80%49) than for CVD (where underreporting and specific diagnosis reporting can be an issue50, although better results are seen for more inclusive, composite categories). Those classified as using NSAID for HOA symptom and pain management may have been using low-dose aspirin for CVD prevention, which could account for the counterintuitive CVD findings. We used stringent definitions of rHOA (through the GOGO criteria) and AUSCAN thresholds that reflect a more severe phenotype and may not be generalizable to milder disease. Finally, while a strength of this community-based study is that findings may be extended to the general population, the lower-income, semirural southern US population of the JoCo OA may limit generalizability.
Overall, DM and CVD were common comorbidities in this cohort, in line with trends in the general population, and having or developing either condition alone or in combination tended to result in a reduced likelihood of symptomatic and functional improvement and an increased likelihood for worsening in both outcomes regarding the hands over time. With an aging population and growing presence of multimorbidity among middle age and older adults, the relationships among common chronic diseases, such as OA, DM, and CVD, are of interest and warrant further research.
Acknowledgment
We thank the staff and participants in the Johnston County Osteoarthritis Project, without whom this work would not be possible.
Footnotes
Funding was provided in part by the Medical Student Training in Aging Research program through NIA 2-T35-AG038047-08, NIAMS P60 AR049465, and P30 AR072580, and CDC U01DP006266. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC.
- Accepted for publication January 29, 2020.