Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Oct 20.
Published in final edited form as: Eur Radiol. 2022 Jan 19;32(6):3944–3953. doi: 10.1007/s00330-021-08471-y

Statin use and MRI subchondral bone marrow lesion worsening in generalized osteoarthritis: longitudinal analysis from Osteoarthritis Initiative data

Bahram Mohajer 1, Ali Guermazi 2, Philip G Conaghan 3, Francis Berenbaum 4, Frank W Roemer 2,5, Arya Haj-Mirzaian 1, Clifton O Bingham 6, Kamyar Moradi 7, Xu Cao 8, Mei Wan 8, Shadpour Demehri 1
PMCID: PMC9583891  NIHMSID: NIHMS1826635  PMID: 35043291

Abstract

Objectives

To determine the association between statin therapy and knee MRI-detected subchondral bone marrow lesion (BML) longitudinal worsening in patients with Heberden’s nodes (HNs) as the hallmark of generalized osteoarthritis (OA) phenotype.

Methods

All participants gave informed consent, and IRB approved HIPAA-compliant protocol. We assessed the worsening in BML volume and number of affected subregions in the Osteoarthritis Initiative (OAI) participants with HNs at baseline clinical examination (HN+), using the semi-quantitative MRI Osteoarthritis Knee Scores at baseline and 24 months. Participants were classified according to baseline BML involvement as “no/minimal” (≤2/14 knee subregions affected and maximum BML score ≤ 1) or “moderate/severe.” Statin users and non-users were selected using 1:1 propensity-score (PS) matching for OA and cardiovascular disease (CVD)–related potential confounding variables. We assessed the association between statin use and increasing BML score and affected subregions using adjusted mixed-effect regression models.

Results

The PS-matched HN+ participants (63% female, aged 63.5 ± 8.5-year-old) with no/minimal and moderate/severe BML cohorts consisted of 332 (166:166, statin users: non-users) and 380 (190:190) knees, respectively. In the HN+ participants with no/minimal BML, statin use was associated with lower odds of both BML score worsening (odds ratio, 95% confidence interval: 0.62, 0.39–0.98) and increased number of affected subregions (0.54, 0.33–0.88). There was no such association in HN− participants or those HN+ participants with baseline moderate/severe BML.

Conclusion

In patients with CVD indications for statin therapy and generalized OA phenotype (HN+), statin use may be protective against the OA-related subchondral bone damage only in the subgroup of participants with no/minimal baseline BML.

Keywords: Magnetic resonance imaging, Bone marrow, Hydroxymethylglutaryl-CoA reductase inhibitors, Osteoarthritis, Knee

Introduction

Knee osteoarthritis (OA) is the most common debilitating disease of the peripheral joints. Despite its high prevalence, to date, no disease-modifying OA drug (DMOAD) has been approved for use in clinical practice [1]. There have been investigations on the potential DMOAD role for statins, a group of first-line lipid-lowering medications. Despite the well-established experimental evidence for the protective effect of statins against knee OA progression and subchondral bone damage in animal models [2], previous observational studies on human OA patients have been inconclusive [3-13]. Such discrepancy could be due to heterogeneous subject selection related to different OA phenotypes, degree of baseline joint structural damage, and most importantly, presence vs. absence of cardiovascular disease (CVD) indications for statin use among participants. While there are controversial reports on the causal relationship between CVD and OA [14-18], previous studies have shown OA is strongly associated with CVDs and CVD risk factors such as obesity and dyslipidemia [15, 19-21]. Therefore, the presence of CVD indications of statins can possibly confound or intermediate statins’ DMOAD assessment.

Experimental studies have shown a protective effect of statins on the subchondral bone [2]. In human OA patients, subchondral bone marrow lesions (BMLs) are known as the imaging hallmark of OA-related subchondral bone damage in MRI examinations [22]. To date, no study has assessed the effect of statins on BMLs, and only one trial on the statins’ DMOAD effect on cartilage loss has been conducted [23, 24]. In this trial, the authors used a 2-year follow-up magnetic resonance imaging (MRI) and included participants who had no CVD indications for statin use and had heterogeneous OA etiologies [23, 24]. While the authors reported that overall statins had no protective effects against OA progression, they stated that statin might reduce cartilage loss only in the subgroup of OA patients with no subchondral BMLs [23, 24]. Moreover, we have previously shown that statin use is associated with decreased radiographic knee OA radiographic progression compared to non-use, only in those OA participants with Heberden’s nodes (HN+) [25]. HNs are bony enlargements of the distal interphalangeal joints (DIPs) detectable in clinical examination and are considered a hallmark of generalized OA phenotype [26-28].

Using the results of the only available clinical trial and recent observational data on HN+ patients, as potential responders to the DMOAD effects of statin, we hypothesized that statins potential DMOAD role might be through their protective effect on early subchondral BML formation and worsening in a distinct subgroup of OA patients with generalized OA (HN+), no/minimal BMLs, and CVD indications for statin use. Using propensity-score (PS) matching for CVD factors and potential confounding by indication (OA and CVD) covariates, we tested this hypothesis in participants of the Osteoarthritis Initiative (OAI) ancillary studies with tailored selection criteria for assessing worsening of MRI-based subchondral bone OA-related damage over a 24-month follow-up.

Materials and methods

Study population

In this study, we used data from the longitudinal multi-center Osteoarthritis Initiative (OAI) study (2004–2015, clinicaltrials.gov identifier: NCT00080171, details can be found at https://nda.nih.gov/oai/). All enrolled patients filled written informed consent and institutional review boards of four OAI collaborating centers have approved the Health Insurance Portability and Accountability Act–compliant protocol of this study. We collected and pooled all previously conducted MRI-based measurements of participants from nested ancillary studies performed inside OAI to assess OA-related subchondral bone damage (Fig. 1). These studies’design and selection criteria are specially tailored to assess MRI-based OA structural damage, worsening in a specific subset from all OAI participants (details are explained in the OAI online repository [29]). Following deletion of duplicate measurements (753 cases between different projects), MRI Osteoarthritis Knee Score (MOAKS) measurements for 1677 knees were included from the following OAI ancillary studies: (1) Foundation for the National Institute of Health (FNIH) Consortium Osteoarthritis Biomarkers Project [30] (473 knees, project no. 22), (2) project no. 30 (125 knees), (3) projects no. 63A-63F (328 knees), (4) pivotal OAI MRI analyses (POMA) study (751 knees) [31]. The same OAI team centrally performed all measurements according to the validated semi-quantitative MOAKS [32].

Fig. 1.

Fig. 1

Open in a new tab

Flowchart of study participants and selection criteria. BML bone marrow lesion, HN Heberden’s node, KL Kellgren-Lawrence, OAI Osteoarthritis Initiative, PS propensity-score

Since it is difficult to assess structural OA damage in patients with advanced knee OA due to “ceiling” effects on the scores, knees with end-stage knee OA on baseline X-ray were excluded, (exclusion #1 in Fig. 1). These consist of knees with replacement surgery (63 knees) and baseline radiographic Kellgren-Lawrence (KL) grade of 4 (302 knees). Moreover, knees without available baseline and 24-month follow-up evaluation of BMLs in the mentioned OAI ancillary studies were excluded (7641 knees, exclusion #2 in Fig. 1). Both knees were included in a minority of participants (N:36, 5% of included knees).

To assess the potential skewness of our sampling and risk of selection bias, we compared the baseline characteristics of OAI participants included in the ancillary studies and the rest of the OAI participants. There was no significant difference in neither of the potential confounders. (Supplementary Table 1).

Assessment of HNs

At the baseline visit, trained OAI nurse staff examined whether HNs on the DIP joints of the 2nd–5th digits and first interphalangeal joint were present by palpation. Participants with at least one HN in either hand were categorized as HN+, whereas participants free of HN in both hands were categorized as HN− and were separately assessed in the sensitivity analysis (sensitivity analysis #1 in Fig. 1).

MRI acquisition and outcome measures

MRI acquisition was performed using 3-T MRI systems (Trio, Siemens Healthcare). Parameters and pulse sequence protocol of OAI MRIs have been previously reported [32]. The validated semi-quantitative MOAKS method was used to assess BMLs at baseline, and follow-up MRIs and features of BML size and number of affected subregions in all 14 anatomical knee joint sub-regions were extracted [32]. BML volume was scored based on the percentage of the total subregion volume occupied as 0: none, 1: < 33%, 2: 33–66%, and 3: > 66% of joint/sub-region volume. To categorize knees according to baseline BML status, we considered both the BML score and the number of affected knee joint subregions. Knees with both criteria of (a) ≤ 2 knee subregions with BMLs and (b) maximum BML score ≤ 1 were considered no/minimal BML involvement. Subsequently, knees with either (a) > 2 knee subregions with BMLs or (b) maximum BML score > 1 were considered with moderate/severe BML involvement. A 24-month BML score worsening was defined as a whole- or within-grade change, where within-grade was defined as a definite visual change while not fulfilling a whole-grade change definition. BML worsening for longitudinal analysis (i.e., outcome to the models) was defined according to previously validated measures [33] as follows: (1) worsening in the number of affected subregions with BML (ranged from improvement to no change, worsening in 1 subregion, and worsening in ≥ 2 subregions), (2) maximum worsening in BML score (ranged from no change, within-grade worsening, to worsening by 1 grade, and worsening by ≥ 2 grades), and (3) worsening in either of BML score (whole or within-grade) or the number of affected subregions (yes/no) [33].

Definition of statin use

According to the OAI protocol, participants were asked to bring their medications at baseline and annual visits. Staff recorded all information on statin type, frequency, and duration of use at each visit, and data were recorded in the OAI Medication Inventory forms (MIFs). To determine the accuracy of selfreported dosage, type, and duration of statin use, in statin users, we extracted and used all available data about the indication of treatment (e.g., primary dyslipidemia, diabetes, heart disease, or cerebrovascular accident), type of statin (including atorvastatin, lovastatin, fluvastatin, simvastatin, pravastatin, and rosuvastatin), and duration of statin use from the OAI MIF dataset. Participants who reported at least one year (equal to 50% of follow-up duration) statin use in OAI MIF forms were considered statin users. Participants who had< 1 year of statin use (two participants in the PS-matched cohorts) or did not report statin use were regarded as statin non-users.

Statistical analysis

Propensity score matching

To minimize the confounding by indication bias, we matched study subcohorts for potential confounders (CVD-related factors: indications of statin use) using baseline clinical characteristics. Potential confounders were investigated using a Direct Acyclic Graph to assess causal inference [34] (Supplementary Fig. 1). The missing data pattern was evaluated, and missing covariate data were imputed. A list of confounding variables and details of the imputation method is presented in the supplementary material.

The matching process was performed using the 1:1 PS-matching method separately in HN+ with no/minimal BMLs and HN+ with moderate/severe BMLs subcohorts; for every knee of statin users, one best-matched knee of the referent (non-users) was selected. We used the nearest neighbor method with a caliper distance of 0.1 calculated with a logistic regression model. We calculated the standardized mean difference (SMD) before and after PS-matching to examine the balance of covariate distribution between the statin users and non-users subcohorts and defined imbalance as an SMD≥0.1.

Regression models

All statistical analyses were separately performed in the HN+ subcohorts with no/minimal BMLs and with moderate/severe BMLs to further assess our hypothesis on statin’s effect on the HN+ statin users with no/minimal baseline BML. We used logistic mixed-effect regression models while considering random intercept for each cluster of matched statin user:non-user and within-subject similarities (due to the inclusion of both knees in a minority (N:36, 5%) of included knees).Statin use was the independent predictor, and BML worsening variables were the dependent outcomes. All models were adjusted for participants’ propensity scores, baseline KL grade, medial joint space narrowing grade, and knees’ BML status.

Sensitivity analysis

We performed the same PS-matched analyses mentioned above on all eligible participants (irrespective of OA phenotype) and the HN− participants to assess whether our results were sensitive to stratification for OA phenotype (sensitivity analysis #1 in Fig. 1). We also evaluated the sensitivity of our results to data imputation with the exclusion of participants with imputed missing data (sensitivity analysis #2 in Fig. 1). Moreover, we assessed sensitivity to PS-matching by performing the analyses on the entire cohort of eligible HN+ OAI participants without PS-matching (sensitivity analysis #3 in Fig. 1). Finally, we evaluated sensitivity to the random exclusion of one of the two knees of participants with both knees included (sensitivity analysis #4).

The open-source R software version 4.0.3 (MASS, haven, survival, Matchlt, mice, lme4, Imer Test, and tableone packages) was used for statistical analysis.

Results

Participants’ characteristics

After the implementation of exclusion criteria and PS-matching, from a total of 9592 knees in the OAI, 332 (statin user: non-user, 166:166) matched knees of HN+ with no/minimal BMLs and 380 (190:190) matched knees of HN+ with moderate/severe BMLs were included in the analysis (Fig. 1). The baseline characteristics of included knees before and after PS-matching are shown in Table 1. The SMD was less than 0.1 for all variables included in the PS-matching model. Participants in all PS-matched cohorts were on average ± standard deviation 63.5 ± 8.5-year-old, were 63% women, and had an average BMI of 29 ± 4.5 kg/m2.

Table 1.

Baseline characteristics of the study population before and after propensity score matching for statin use according to baseline BML in MRI

HN+ participant with no/minimal BMLs HN+ participant with moderate/severe BMLs
Characteristic Before matching Propensity score-matched Before matching Propensity score-matched
Statin (−) Statin (+) SMD Statin (−) Statin (+) SMD Statin (−) Statin (+) SMD Statin (−) Statin (+) SMD
No. of knees 339 175 166 166 386 203 190 190
Variables in the P.S matching (Potential confounders)
Age (year) [mean (SD)] 60.99 (8.69) 63.82 (7.87) 0.342 63.66 (8.41) 63.61 (7.93) 0.006 62.91 (8.65 64.91 (7.75) 0.244 63.94 (7.54) 64.75 (7.80) 0.098
No. of women [N (%)] 239 (70.5) 109 (62.3) 0.175 110 (66.3) 104 (62.7) 0.076 251 (65.0) 125 (61.6) 0.072 120 (63.2) 118(62.1) 0.022
BMI (kg/m 2 ) [mean (SD)] 28.10 (4.69) 28.40 (3.84) 0.069 28.15 (3.94) 28.31 (3.91) 0.04 28.92 (4.46) 30.05 (4.19) 0.261 29.50 (4.67) 29.72 (3.97) 0.049
Statin CVD indications except dyslipidemia[N(%)]♦ 76 (22.4) 52 (29.7) 0.167 48 (28.9) 50 (30.1) 0.026 119 (30.8)66 (17.1) 89 (43.8) 0.272 76 (40.0) 78 (41.1) 0.021
Alcohol use, ≥ 1/week [N(%)] 167 (49.3) 75 (42.9) 0.129 76 (45.8) 75 (45.2) 0.012 174(45.1) 83 (40.9) 0.085 78 (41.1) 81 (42.6) 0.032
Smoking, current or past [N (%)] 151 (44.5) 92 (52.6) 0.161 84 (50.6) 87 (52.4) 0.036 170 (44.0) 101 (49.8) 0.115 90 (47.4) 91 (47.9) 0.011
PASE score [mean (SD)] 166.32 (79.40) 155.45 (70.78) 0.145 156.54 (74.01) 155.16 (72.30) 0.019 164.61 (82.81) 150.55 (72.49) 0.181 152.93 (78.22) 152.84 (73.36) 0.001
Race, non-white [N (%)]† 42 (12.4) 19 (10.9) 0.048 17 (10.2) 17 (10.2) 0.001 66 (17.1) 34 (16.7) 0.009 32 (16.8) 32 (16.8) 0.001
Other variables not in the P.S matching
History of knee injury, [N(%] 80 (23.6) 36 (20.6) 0.07 37 (22.3) 35 (21.1) 0.03 121 (31.3) 72 (35.5) 0.09 59 (31.1) 65 (34.2) 0.07
Statin type – – – –
Atorvastatin – 79 (60.3) – 73 (59.3) – 82 (48.2) – 77 (49.0)
Fluvastatin – 2(1.5) – 2(1.6) – 3(1.8) – 2(1.3)
Lovastatin – 6 (4.6) – 6(4.9) – 9(5.3) – 7 (4.5)
Pravastatin – 9 (6.9) – 8 (6.5) – 14 (8.2) – 14 (8.9)
Rosuvastatin – 6 (4.6) – 6 (4.9) – 7(4.1) – 6(3.8)
Simvastatin – 29 (22.1) – 28 (22.8) – 55 (32.4) – 51 (32.5)
Statin use duration, years, [mean (SD)] 0.00 (0.00) 3.92 (2.05) 2.7 0.00 (0.00) 3.90 (2.06) 2.68 0.00 (0.00) 4.23 (2.06) 2.9 0.00 (0.00) 4.18(2.08) 2.85
Number of affected subregions with BML 0.07 0.05 0.2 0.32
0 206 (60.8) 101 (57.7) 95 (57.2) 94 (56.6) 75 (19.4) 39 (19.2) 37 (19.5) 37 (19.5)
1 83 (24.5) 44 (25.1) 45 (27.1) 43 (25.9) 24 (6.2) 18 (8.9) 7 (3.7) 17 (8.9)
2 50 (14.7) 30 (17.1) 26 (15.7) 29 (17.5) 51 (13.2) 34 (16.7) 22(11.6) 34 (17.9)
3 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 99 (25.6) 55 (27.1) 56 (29.5) 51 (26.8)
4 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 64 (16.6) 29 (14.3) 32 (16.8) 25 (13.2)
≥5 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 73 (18.9) 28 (13.8) 36(18.9) 26 (13.7)
Maximum BML grade in knee 0.16 0.03 0.02 0.09
0 148 (43.7) 63 (36.0) 63 (38.0) 61 (36.7) 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0)
1 191 (56.3) 112(64.0) 103 (62.0) 105 (63.3) 58 (15.0) 30 (14.8) 34 (17.9) 28 (14.7)
2 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 217 (56.2) 116(57.1) 107 (56.3) 109 (57.4)
3 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 111(28.8) 57 (28.1) 49 (25.8) 53 (27.9)
Baseline KL grade 0.21 0.235 0.062 0.166
Grade 0 111(32.7) 49(28.0) 53 (31.9) 45 (27.1) 54 (14.0) 28 (13.8) 22(11.6) 28 (14.7)
Grade 1 102 (30.1) 68 (38.9) 50(30.1) 64 (38.6) 92 (23.8) 50 (24.6) 43 (22.6) 48 (25.3)
Grade 2 94 (27.7) 39 (22.3) 49 (29.5) 38 (22.9) 137 (35.5) 76 (37.4) 65 (34.2) 67 (35.3)
Grade 3 32 (9.4) 19 (10.9) 14 (8.4) 19(11.4) 103 (26.7) 49(24.1) 60(31.6) 47 (24.7)
Open in a new tab

Data are presented as numbers of knees. Statin (+) and statin (−) corresponds to statin users and non-users, respectively. Knees with both ≤2 knee subregions with BMLs and maximum BML score ≤ 1 were regarded with no/minimal BML involvement, while knees either having > 2 knee subregions with BMLs or maximum BML score > 1 were considered with moderate/severe BML involvement. BMI body mass index, BML Bone marrow lesion, HN Heberden’s node, PASE physical activity scale for the elderly, SMD standardized mean difference, SD standard deviation, N number of knees

A significant difference for SMD was defined as ≥0.1

♦

Statin CVD indications except dyslipidemia were indicated as the presence of either history of coronary artery disease, cerebrovascular accident, diabetes (any stage of diabetes vs. no medical history of diabetes), or hypertension in clinical examination systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥ 90 mm Hg at OAI visit clinical examination)

†

Race of participants was categorized as white and non-white considering the small number of participants in each non-white race group

Outcome measures

In HN+ with no/minimal baseline BML subcohort, statin use was associated with lower odds of an increasing number of affected subregions with BML (odds ratio, 95% confidence interval: 0.54, 0.33–0.88), BML score worsening (0.62, 0.39–0.98), and worsening in either BML score or the number of affected subregions (0.60, 0.37–0.99) (Table 2). There was no such association in the HN+ with moderate/severe baseline BML subcohort (worsening in the number of affected subregions: 1.04, 0.70–1.53, BML score worsening: 0.96, 0.65–1.42, and worsening in either BML score or the number of affected subregions: 0.85, 0.50–1.47) (Table 2).

Table 2.

Longitudinal 24-month assessment of subchondral BML worsening in MRI between propensity-score matched HN+ statin users vs. non-users, according to BML involvement of the knee joint in the baseline visit MRI examination

HN+
Statin user: non-user No/minimal BML in baseline MRI Moderate/severe BML
in baseline MRI
N: 332 (166:166) N: 380(190:190)
Worsening in number of affected subregions with BML 0.54 (0.33–0.88), p:0.015 1.04 (0.7–1.53), p:0.859
Maximum worsening in BML score 0.62 (0.39–0.98), p:0.041 0.96 (0.65–1.42), p:0.841
Worsening in BML score or number of affected subregions 0.60 (0.37–0.99), p:0.044 0.85 (0.50–1.47), p:0.566
Open in a new tab

Participants in the HN+ subcohort were separately matched for possible confounders with the 1:1 PS matching method. Longitudinal measures of BML worsening were compared between matched statin users: non-users using logistic mixed-effect linear models. Previously validated longitudinal 24-month BML dependent variables (i.e., outcome to the models) included (1) worsening in the number of affected subregions with BML (ranged from Improvement to no change, worsening in 1 subregion, and worsening in ≥2 subregions), (2) maximum worsening in BML score (ranged from no change, within-grade worsening, to worsening by 1 grade, and worsening by ≥2 grades), and (3) worsening in either of BML score (whole or within-grade) or the number of affected subregions (yes/no). All models were adjusted for participants propensity score, baseline Kellgren-Lawrence grade, medial Joint Space Narrowing (JSN) grade, and BML status (two variables of (1) number of affected subregions affected by BMLs and (2) max BML score in the joint) while considering random intercept for each cluster of matched statin user: non-user and random intercept considering within-subject similarities (due to the inclusion of both knees in a minority of participants, 5%) where the knee is nested within participant ID. All analyses were categorized according to baseline BML involvement in MRI. Knees with both ≤2 knee subregions with BMLs and maximum BML score ≤1 were regarded with no/minimal BML involvement, while knees either having >2 knee subregions with BMLs or maximum BML score > 1 were considered with moderate/severe BML involvement. BML bone marrow lesions, HN Heberden’s node

Sensitivity analysis

Our sensitivity analysis showed that without stratification for OA phenotype (all HN+ and HN−s) or in HN− subcohort, there was no association between statin use and BML worsening. (Supplementary Tables 2 and 3) Furthermore, our results were not sensitive to using the PS-matching method, data imputation, or random exclusion of one knee of participants with both knees included (Supplementary Table 2).

Discussion

Using the available data from our previously published paper on HN+ participants and the recent trial, we have tested the hypothesis that statin use is associated with reduced BML worsening over two years in the knee joint, only in a specific OA phenotype with HNs, no/minimal BMLs, and with CVD indications for statin use. Our finding suggests that the protective effect of statins against OA-related subchondral bone damage, which is not seen in all OA patients and is exclusive to patients with HNs (as the hallmark of a generalized OA phenotype) [25] may be associated with a reduction in early OA-related subchondral damage.

The current data on statins’ effects against OA-related outcomes is controversial and limited to observational studies [3-13]. The reasons for the overall inconclusive results of previous observational clinical studies could be due to the inclusion of heterogeneous OA populations in terms of OA phenotype, degree of baseline structural damage in the joint, and underlying comorbidities (e.g., CVDs) that may mediate or confound the potential DMOAD role of statins. We, therefore, carefully formed our hypothesis and selected participants using findings of our previous observational study on HN+ patients (generalized OA) [25] and the only conducted clinical trial (minimal/no BML) [23, 24], while trying to address potential limitations of these studies (e.g., excluding patients with CVD indications for statin use in the trial). Considering the inclusion of patients with generalized OA, a large body of literature supports HNs as the hallmark of generalized OA and a strong predictor of knee OA progression [26, 35]. Previous studies have shown generalized OA and HNs in DIPs are also strongly associated with CVD risk factors such as elevated serum cholesterol and lipid dysregulation [36]. Moreover, we have previously shown HN+ OA patients have 40% higher odds of OA MRI-detected subchondral damage during 24-months of follow-up compared to HN− patients [37] Valdes et al., using a cross-sectional design, demonstrated a significant association between statin use and less severe hip and knee OA—assessed by the Kellgren-Lawrence grading system—exclusively in patients with generalized OA [7]. We have recently shown that statin use is associated with a 46% reduced risk of radiographic progression of OA over 8 years compared to no use, only in HN+ patients and not HN−s [25].

On the other hand, from the only clinical trial on statins with a 2-year follow-up [23, 24], the authors reported a protective effect for statins on OA progression, but only in participants with no baseline subchondral BMLs, another finding that helped to form our hypothesis and selecting participants. However, according to the trial inclusion criteria, authors excluded patients with CVD indications of statin use, and participants used statins purely for OA progression [23, 24]. This may have resulted in excluding the population who could benefit from statins’ effects on subchondral bone. More importantly, the authors did not consider OA phenotypes (e.g., generalized OA) in the subject selection. Our sensitivity analysis showed that when assessing all OA patients irrespective of their phenotype (both HN+ and HN−), similar to this trial, we observed no protective association with statin use.

While showing beneficial effects of statins in an OA population who already have statin use indication may first seem only incremental in clinical practice, a considerable beneficial epidemiologic impact of DMOAD role for statins can be expected in two distinct patient populations. The first population is current statin users for CVD and its risk factors. Statins are among the most prescribed medications in the elderly, mainly indicated for dyslipidemia and other CVD risk factors [38]. One of the main challenges for statin use is the disappointing long-term adherence rate of as low as 25% [39] due to reasons like perceived lack of efficacy of statins and subjective musculoskeletal pain or its related subjective concerns (also known as statin-associated muscle symptoms or SAMS) [40]. Furthermore, older patients [41] (also more affected by OA) and those with debilitating comorbidities like OA [42, 43] are among the groups with the least adherence to statins [44, 45]. The second population who benefit is generalized OA statin “non-users” with a CVD indication for statin use. Reports show that a third of the adults in developed countries like the USA meet statin CVD indications, but nearly half of this population have never initiated statin use [46]. If the potential DMOAD role for statins is proven, it can improve both initiation of and adherence to one of the world’s most commonly prescribed medications [44, 45].

As for the strengths of the current study, we studied a hypothesis-driven selected large sample of PS-matched participants from the validated OAI cohort. Also, we used MOAKS scorings which have been shown to closely correlate with pain and structural damage or the progression of OA [33]. Previous studies on the OAI data reported intra- and inter-observer reliability of 90% for longitudinal BML MOAKS measurement, suggesting a low risk of measurement error for influencing outcome results [47]. Moreover, we uniquely selected and PS-matched participants based on previous evidence and performed several stratification and sensitivity analyses to OA phenotype (HN−, moderate/severe BMLs) and selected methods (PS-matching, missing data imputation, the inclusion of both knees) to assess the robustness of our results.

However, our study has several limitations. First, we lacked precise data on the duration, dosage, and intensity of statin use. OAI examiners confirmed the prescription for statins according to medications participants brought with them during visits. This approach may not be as valid as the exact pill count and cannot be used for exploring the dose-dependent effects of statins, but it may be more reliable than a self-report of medication use. A similar approach has been implemented in previous OAI studies [8, 13, 25], and it has been shown that these measures of statin use are relatively accurate [48]. Second, we have included all statin users with different statin use duration before the baseline visit (both prevalent and incident users). This will increase the risk of Neyman bias in our results, which is a selection bias in which very sick/healthy participants (because of chronic disease) are excluded from enrolment [49]. Third, our defined subcohorts were not pre-specified in the OAI data collection process because of the retrospective analysis of the prospectively collected data. We were limited by including participants with available MRI scorings from previously conducted nested case–control studies within OAI (e.g., FNIH, POMA), which have specific inclusion-exclusion criteria. We tried to tailor our study sample to address this limitation using detailed selection criteria and the PS-matching method, and we assessed the sensitivity of our results to using the PS-matching method. Forth, in assessing CVD statin indications, lipid profile was not available in the OAI dataset, and dyslipidemia is among the most common indications of statin prescription for primary CVD prevention [50]. While we tried to match our participants according to other statin CVD statin indications and the majority (> 70%) of our participants had statin indications (all statin users, 40% of non-users according to PS-matching results), it is not possible to thoroughly address this issue in studies with an observational design where exposure (statin use) was not considered in selection criteria. Finally, we have not assessed SAMS and muscle strength and quality in this study, a matter that can potentially complicate the implementation of statin DMOAD role in routine clinical practice. Given the high prevalence of SAMS, detecting any deterioration of muscle quality will raise a critical concern for statins' DMOAD role in clinical practice, a matter left for future studies.

In conclusion, our results suggest that statin use may be protective against BML worsening only in a specific OA phenotype with HNs and no/minimal baseline BMLs and with CVD statin indications, which is in line with the recent observational data [25] and the only available clinical trial [23, 24]. While our exploratory study results cannot be directly translated to clinical use, future studies focusing on the repurposing of widely available statins as DMOADs [51] with proper patient selection may produce clinical and potential cost-saving benefits compared to designing new DMOAD compounds.

Supplementary Material

Supplementary material

Key Points.

  • Statin use may reduce the risk of subchondral bone damage in specific osteoarthritis patients with a generalized phenotype, minimal subchondral bone damage, and cardiovascular statin indications.

Acknowledgements

The authors would like to thank the participants and staff involved in OAI, FNIH, and POMA projects. Several grants and direct or in-kind contributions provide the publicly available data from the FNIH OA Biomarkers Consortium Project, including Abb Vie, Amgen, Arthritis Foundation, Artialis; Bioiberica, Bio Vendor, DePuy, Flexion Therapeutics, GSK, IBEX, IDS, Merck Serono, Quidel, Rottapharm I Madaus, Sanofi, Stryker, the Pivotal OAI MRI Analyses (POMA) study, NIH HHSN2682010000 21C, and the Osteoarthritis Research Society International. The OAI is a public-private partnership comprised of five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health. This manuscript was prepared using an OAI public use data set and does not necessarily reflect the opinions or views of the OAI investigators, the NIH, or the private funding partners. PGC is supported in part through the UK NIHR Leeds Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the funding partners and sponsors.

Funding

This research was supported by the NIH National Institute of Aging (NIA) under Award Number P01AG066603 and NIH National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) under Award Number R01AR079620-01.

Abbreviations

BML

Bone marrow lesion

CVD

Cardiovascular disease

DMOAD

Disease-modifying osteoarthritis drug

HN

Heberden’s node

MIF

Medication inventory form

MOAKS

MRI Osteoarthritis Knee Score

OA

Osteoarthritis

OAI

Osteoarthritis Initiative

PS

Propensity-score

SAMS

Statin-associated muscle symptoms

SMD

Standardized mean difference

Footnotes

Conflict of interest FWR is chief marketing officer and shareholder of Boston Imaging Core Lab (BICL), LLC, and consultant to Calibr – California Institute of Biomedical Research and Grünenthal GmbH. AG is a shareholder of BICL and consultant to Pfizer, TissueGene, Merck Serono, Novartis, Regeneron, and AstraZeneca. FB reports personal fees from AstraZeneca, Boehringer, Bone Therapeutics, CellProthera, Expanscience, Galapagos, Gilead, Grunenthal, GSK, Eli Lilly, Merck Sereno, MSD, Nordic, Nordic Bioscience, Novartis, Pfizer, Roche, Sandoz, Sanofi, Servier, UCB, Peptinov, 4P Pharma, grants from TRB Chemedica, non-financial support from 4 Moving Biotech, outside the submitted work. The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00330-021-08471-y.

Guarantor The scientific guarantor of this publication is Dr. Shadpour Demehri.

Statistics and biometry No complex statistical methods were necessary for this paper.

Informed consent Written informed consent was obtained from all subjects (patients) in the osteoarthritis initiative (OAI) study. Written informed consent was not directly required for this study, since we have used the data of the OAI study.

Ethical approval Institutional Review Board approval was obtained. The study has received ethics board approval by the institutional review board at the University of California, San Francisco (OAI Coordinating Center; Approval Number: 10-00532), and all enrolled subjects gave informed consent.

Study subjects or cohorts overlap Osteoarthritis Initiative (OAI) is a well-known publicly available dataset. Some study subjects or cohorts have been previously reported in studies published using the OAI and/or FNIH study. The list of studies on the OAI dataset can be found in the online address of https://nda.nih.gov/oai/publications.

Methodology
  • retrospective
  • case-control
  • multicenter study

References

  • 1.Bijlsma JWJ, Berenbaum F, Lafeber FPJG (2011) Osteoarthritis: an update with relevance for clinical practice. Lancet 377:2115–2126 [DOI] [PubMed] [Google Scholar]
  • 2.Famaghi S, Prasadam I, Cai G et al. (2017) Protective effects of mitochondria-targeted antioxidants and statins on cholesterolinduced osteoarthritis. FASEB J 31:356–367 [DOI] [PubMed] [Google Scholar]
  • 3.Kadam UT, Blagojevic M, Belcher J (2013) Statin use and clinical osteoarthritis in the general population: a longitudinal study. I Gen Intern Med 28:943–949 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Makris UE, Alvarez CA, Mortensen EM, Mansi IA (2018) Association of statin use with increased risk of musculoskeletal conditions: a retrospective cohort study. Drug Saf 41:939–950 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Clockaerts S, Van Osch GJVM, Bastiaansen-Jenniskens YM et al. (2012) Statin use is associated with reduced incidence and progression of knee osteoarthritis in the Rotterdam study. Ann Rheum Dis 71:642–647 [DOI] [PubMed] [Google Scholar]
  • 6.Wang J, Dong J, Yang J, Wang Y, Liu J (2020) Association between statin use and incidence or progression of osteoarthritis: meta-analysis of observational studies. Osteoarthritis Cartilage 28:1170–1179 [DOI] [PubMed] [Google Scholar]
  • 7.Valdes AM, Zhang W, Muir K, Maciewicz RA, Doherty S, Doherty M (2014) Use of statins is associated with a lower prevalence of generalised osteoarthritis. Ann Rheum Dis 73:943–945 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Riddle DL, Moxley G, Dumenci L (2013) Associations between Statin use and changes in pain, function and structural progression: a longitudinal study of persons with knee osteoarthritis. Ann Rheum Dis 72:196–203 [DOI] [PubMed] [Google Scholar]
  • 9.Michaëlsson K, Lohmander LS, Turkiewicz A, Wolk A, Nilsson P, Englund M (2017) Association between statin use and consultation or surgery for osteoarthritis of the hip or knee: a pooled analysis of four cohort studies. Osteoarthritis Cartilage 25:1804–1813 [DOI] [PubMed] [Google Scholar]
  • 10.Mansi IA, Mortensen EM, Pugh MJ, Wegner M, Frei CR (2013) Incidence of musculoskeletal and neoplastic diseases in patients on statin therapy: results of a retrospective cohort analysis. Am J Med Sci 345:343–348 [DOI] [PubMed] [Google Scholar]
  • 11.Eymard F, Parsons C, Edwards MH et al. (2017) Statin use and knee osteoarthritis progression: results from a post-hoc analysis of the SEKOIA trial. Joint Bone Spine 85:609–614 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Chodick G, Amital H, Shalem Y et al. (2010) Persistence with statins and onset of rheumatoid arthritis: a population-based cohort study. PLoS Med 7:e1000336–e1000336 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Veronese N, Koyanagi A, Stubbs B et al. (2019) Statin use and knee osteoarthritis outcomes: a longitudinal cohort study. Arthritis Care Res (Hoboken) 71:1052–1058 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Kuusalo L, Felson DT, Wang N et al. (2021) Metabolic osteoarthritis - relation of diabetes and cardiovascular disease with knee osteoarthritis. Osteoarthritis Cartilage 29:230–234 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Hall AJ, Stubbs B, Mamas MA, Myint PK, Smith TO (2016) Association between osteoarthritis and cardiovascular disease: systematic review and meta-analysis. Eur J Prev Cardiol 23:938–946 [DOI] [PubMed] [Google Scholar]
  • 16.Corti MC, Rigon C (2003) Epidemiology of osteoarthritis: prevalence, risk factors and functional impact. Aging Clin Exp Res 15:359–363 [DOI] [PubMed] [Google Scholar]
  • 17.Fernandes GS, Valdes AM (2015) Cardiovascular disease and osteoarthritis: common pathways and patient outcomes. Eur J Clin Invest 45:405–414 [DOI] [PubMed] [Google Scholar]
  • 18.Ong KL, Wu BJ, Cheung BMY, Barter PJ, Rye K-A (2013) Arthritis: its prevalence, risk factors, and association with cardiovascular diseases in the United States, 1999 to 2008. Ann Epidemiol 23:80–86 [DOI] [PubMed] [Google Scholar]
  • 19.Plumb MS, Aspden RM (2004) High levels of fat and (n-6) fatty acids in cancellous bone in osteoarthritis. Lipids Health Dis 3:12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Findlay DM (2007) Vascular pathology and osteoarthritis. Rheumatology 46:1763–1768 [DOI] [PubMed] [Google Scholar]
  • 21.Kadam UT (2004) Clinical comorbidity in patients with osteoarthritis: a case-control study of general practice consulters in England and Wales. Ann Rheum Dis 63:408–414 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Dore D, Quinn S, Ding C et al. (2010) Natural history and clinical significance of MRI-detected bone marrow lesions at the knee: a prospective study in community dwelling older adults. Arthritis Res Ther 12:R223. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Wang Y, Jones G, Hill C et al. (2020) Effect of atorvastatin on knee structure and pain in patients with knee osteoarthritis: a randomised controlled trial. Osteoarthritis Cartilage 28:S478–S479 [Google Scholar]
  • 24.Wang Y, Jones G, Hill C et al. (2021) Effect of atorvastatin on knee cartilage volume in patients with symptomatic knee osteoarthritis: results from a randomised placebo-controlled trial. Arthritis Rheumatol. 10.1002/art.41760 [DOI] [PubMed] [Google Scholar]
  • 25.Haj-Mirzaian A, Mohajer B, Guermazi A et al. (2019) Statin use and knee osteoarthritis outcome measures according to the presence of Heberden nodes: results from the Osteoarthritis Initiative. Radiology 293:190557. [DOI] [PubMed] [Google Scholar]
  • 26.Kellgren JH, Moore R (1952) Generalized osteoarthritis and Heberden’s nodes. BMJ 1:181–187 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Irlenbusch U, Dominick G (2006) Investigations in generalized osteoarthritis. Part 2: Special histological features in generalized osteoarthritis (histological investigations in Heberden’s nodes using a histological score). Osteoarthritis Cartilage 14:428–434 [DOI] [PubMed] [Google Scholar]
  • 28.Nelson AE, Smith MW, Golightly YM, Jordan JM (2014) “Generalized osteoarthritis”: a systematic review. Semin Arthritis Rheum 43:713–720 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.(2016) Overview and Description of Central Image Assessments. Available via https://nda.nih.gov/binaries/content/documents/ndacms/resources/oai/oai-image-assessment-overview/oai-image-assessment-overview/ndacms:resource. Accessed 04/25/2021 2021
  • 30.Aboulenain S, Saber AY (2021) Primary OsteoarthritisStatPearls. StatPearls Publishing. Copyright © 2021, StatPearls Publishing LLC., Treasure Island (FL) [Google Scholar]
  • 31.Hoaglund FT (2013) Primary osteoarthritis of the hip: a genetic disease caused by European genetic variants. J Bone Joint Surg Am 95:463–468 [DOI] [PubMed] [Google Scholar]
  • 32.Hunter DJ, Guermazi A, Lo GH et al. (2011) Evolution of semiquantitative whole joint assessment of knee OA: MOAKS (MRI Osteoarthritis Knee Score). Osteoarthritis Cartilage 19:990–1002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Collins JE, Losina E, Nevitt MC et al. (2016) Semiquantitative imaging biomarkers of knee osteoarthritis progression: data From the Foundation for the National Institutes of Health Osteoarthritis Biomarkers Consortium. Arthritis Rheumatol 68:2422–2431 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Schisterman EF, Cole SR, Platt RW (2009) Overadjustment Bias and Unnecessary Adjustment in Epidemiologic Studies. Epidemiology 20:488–495 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Blagojevic M, Jinks C, Jeffery A, Jordan KP (2010) Risk factors for onset of osteoarthritis of the knee in older adults: a systematic review and meta-analysis. Osteoarthritis Cartilage 18:24–33 [DOI] [PubMed] [Google Scholar]
  • 36.Kovari E, Kaposi A, Bekes G et al. (2020) Comorbidity clusters in generalized osteoarthritis among female patients: a cross-sectional study. Semin Arthritis Rheum 50:183–191 [DOI] [PubMed] [Google Scholar]
  • 37.Haj-Mirzaian A, Mohajer B, Guermazi A et al. (2019) Heberden’s nodes and knee osteoarthritis–related osseous structural damage: exploratory study from the osteoarthritis initiative. Arthritis Rheumatol 71:935–940 [DOI] [PubMed] [Google Scholar]
  • 38.Smith DA, Galin I (2006) Statin therapy for native and peri-interventional coronary heart disease. Curr Mol Med 6:589–602 [DOI] [PubMed] [Google Scholar]
  • 39.Jackevicius CA, Mamdani M, Tu JV (2002) Adherence with statin therapy in elderly patients with and without acute coronary syndromes. ACC Curr J Rev 11:36. [DOI] [PubMed] [Google Scholar]
  • 40.Wei MY, Ito MK, Cohen JD, Brinton EA, Jacobson TA (2013) Predictors of statin adherence, switching, and discontinuation in the USAGE survey: understanding the use of statins in America and gaps in patient education. J Clin Lipidol 7:472–483 [DOI] [PubMed] [Google Scholar]
  • 41.Warren JR, Falster MO, Fox D, Jorm L (2013) Factors influencing adherence in long-term use of statins. Pharmacoepidemiol Drug Saf 22:1298–1307 [DOI] [PubMed] [Google Scholar]
  • 42.Mauskop A, Borden WB (2011) Predictors of Statin Adherence. Curr Cardiol Rep 13:553. [DOI] [PubMed] [Google Scholar]
  • 43.Parkinson L, Waters DL, Franck L (2017) Systematic review of the impact of osteoarthritis on health outcomes for comorbid disease in older people. Osteoarthritis Cartilage 25:1751–1770 [DOI] [PubMed] [Google Scholar]
  • 44.Mann DM, Allegrante JP, Natarajan S, Halm EA, Charlson M (2007) Predictors of Adherence to Statins for Primary Prevention. Cardiovasc Drugs Ther 21:311–316 [DOI] [PubMed] [Google Scholar]
  • 45.Vinker S, Shani M, Baevsky T, Elhayany A (2008) Adherence with statins over 8 years in a usual care setting. Am J Manag Care 14:388–392 [PubMed] [Google Scholar]
  • 46.Mercado C, DeSimone AK, Odom E et al. (2015) Prevalence of cholesterol treatment eligibility and medication use among adults—United States, 2005–2012. 64:1305–1311 [DOI] [PubMed] [Google Scholar]
  • 47.Roemer FW, Guermazi A, Collins JE et al. (2016) Semi-quantitative MRI biomarkers of knee osteoarthritis progression in the FNIH biomarkers consortium cohort – Methodologic aspects and definition of change. BMC Musculoskelet Disord 17:466. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Boudreau DM, Daling JR, Malone KE, Gardner JS, Blough DK, Heckbert SR (2004) A validation study of patient interview data and pharmacy records for antihypertensive, statin, and antidepressant medication use among older women. Am J Epidemiol 159:308–317 [DOI] [PubMed] [Google Scholar]
  • 49.Danaei G, Tavakkoli M, Hernan MA (2012) Bias in observational studies of prevalent users: lessons for comparative effectiveness research from a meta-analysis of statins. Am J Epidemiol 175:250–262 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Force USPST (2016) Statin use for the primary prevention of cardiovascular disease in adults: US Preventive Services Task Force recommendation statement. JAMA 316:1997–2007 [DOI] [PubMed] [Google Scholar]
  • 51.(2015) New Indications and a Sense of (Re)purpose. EBioMedicine 2:1257–1258 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary material
NIHMS1826635-supplement-Supplementary_material.docx (301.1KB, docx)

RESOURCES