WARFARIN USE AND RISK OF KNEE AND HIP REPLACEMENTS

Abstract

Objectives

Identification of modifiable risk factors and treatments for osteoarthritis (OA) is needed. Warfarin, a vitamin K antagonist, causes fetal and animal model skeletal abnormalities. While vitamin K insufficiency has been associated with OA, whether warfarin is also detrimental to OA is not known. We therefore evaluated the relation of warfarin use to risk of knee and hip replacements (KR, HR) as a reflection of end-stage OA.

Methods

We conducted a nested case-control study using a UK general practitioner electronic medical records database. We identified cases of KR or HR from among adults with atrial fibrillation newly prescribed either warfarin or direct oral anticoagulants (DOAC). Cases were matched with 4 controls by age and sex. We assessed the relation of warfarin compared with DOAC use to risk of joint replacement using conditional logistic regression. We also evaluated different durations of warfarin use.

Results

We identified 857 subjects with KR or HR (cases), of whom 64.6% were warfarin users, and 3428 matched controls, of whom 56.1% were warfarin users (mean age 75, 47% female). Warfarin users had a 1.59 times higher risk of joint replacement than DOAC users (Adjusted OR 1.59 (95% CI 1.31, 1.92). Longer duration of warfarin use was associated with higher risk of joint replacement compared with <1 year of warfarin use.

Conclusion

Warfarin, a vitamin K antagonist, was associated with greater risk of KR and HR (an indicator for end-stage knee OA) than DOAC use, supporting the importance of adequate vitamin K functioning in limiting OA progression.

INTRODUCTION

Warfarin is a commonly prescribed anticoagulant that is known to have adverse effects on the skeletal system in the context of human fetal embryopathy and in rat models characterized by abnormal skeletal mineralization.[1–4] These effects could have implications for osteoarthritis (OA), the most common form of arthritis for which no effective treatments exist. As such, identifying modifiable risk factors remains a high priority.

Warfarin’s anticoagulant effects occur through inhibition of vitamin K’s functioning.[5] Vitamin K, in turn, is an essential co-factor in the post-translational gamma carboxylation of Gla proteins, a step required for these proteins to be functional.[6] Gla proteins play an important role not only in blood coagulation, but also in the bone and cartilage, including matrix Gla protein (MGP), osteocalcin and Gas-6.[7–9] Thus, warfarin’s inhibition of vitamin K leads to inadequate functioning of Gla proteins. Low vitamin K status has been associated with both incidence and progression of knee OA in observational studies.[10–13] Further, in a randomized controlled trial of vitamin K supplementation versus placebo, those with insufficient vitamin K at baseline had trends towards less joint space narrowing on hand radiographs.[14]

Taken together, these data highlight the potential detrimental effects of warfarin via vitamin K antagonism on joint tissues that could contribute to OA pathology. We therefore sought to determine the relation of warfarin use to risk of knee and hip replacements (KR, HR), as a reflection of end-stage OA, in a large population-based cohort.

METHODS

Study Design

We performed a nested case-control study using data from the IQVIA Medical Research Data (IMRD; incorporating The Health Improvement Network (THIN)). IMRD is a general practitioner (GP) electronic medical records database from the United Kingdom (UK) that is representative of the general UK population. This database has been validated for use in pharmacoepidemiologic research.[15]

The nested case-control study was assembled from among a cohort of adults with atrial fibrillation, a common indication for long-term anticoagulation, to minimize confounding by indication. Because atrial fibrillation can be managed with warfarin or direct oral anticoagulants (DOACs), which do not antagonize vitamin K,[16] we utilized an active comparator approach to further minimize confounding by indication. Since DOACs were first introduced to the market in the UK in 2008, we identified eligible study participants from 2009 onwards to allow time for market uptake. Study entry criteria included adults aged between 40 and 89 with atrial fibrillation who were enrolled with a GP for at least 1 year. Participants were further required to be incident warfarin or DOAC users defined as those who were newly prescribed an anticoagulant after 2009, having ≥1 prescription after study entry and also within 1 year prior to the index date (defined below) to ensure a relevant timeframe of use. From among this cohort, we identified cases as those with KR or HR between 2014–2018. The index date for cases was defined as the date of surgery. Each case was matched with 4 controls by age and sex; if more than 4 controls were eligible for matching, the 4 controls were selected randomly. The matched controls were assigned the same index date as that of their matching case’s surgery date (Figure 1).

Figure 1:

Study design and timeline

We excluded participants with KR or HR prior to 2014, those with warfarin or DOAC use prior to study entry (criteria defined above) and those who used both warfarin and DOAC within 1 year prior to the index date. We also excluded those with high-risk cancer (esophageal, gastric, pancreatic and metastatic cancer), body mass index (BMI) > 40 kg/m², joint infection, and oxygen therapy as these are severe comorbidities that would limit surgical candidacy.

Analytic Approach

For our primary analysis, we assessed the relation of warfarin use compared with DOAC use, both within 1 year prior to the index date, to risk of KR and HR. Because the biologic effects of warfarin may become evident only after a period of use, in a secondary analysis, we assessed the relation of duration of warfarin use to risk of KR and HR, defined as ≥4 years, 2-<4 years, and 1-<2 years, compared with warfarin use of <1 year prior to the index date. Duration of use was calculated based upon the sum of each prescription duration between study entry and the index date.

We considered the following potential confounders for adjustment in our models: BMI, renal disease, severe liver disease, prior gastrointestinal bleeding, prior intracranial hemorrhage, mitral stenosis, presence of prosthetic heart valve, prior falls, cancer, chronic obstructive pulmonary disease (COPD), dementia or cognitive impairment, diabetes, heart failure, hypertension, hyperlipidemia, ischemic heart disease (IHD), stroke, venous thromboembolism, medication use (anti-hypertensives, oral hypoglycemic medications, insulin, lipid lowering medications, non-steroidal anti-inflammatory drugs (NSAIDs) and paracetamol), GP visits and hospitalizations. Confounders were assessed by read codes for medical conditions and with prescription records for medications. Of these, severe liver disease, prior intracranial hemorrhage, mitral stenosis, and presence of prosthetic heart valve had a prevalence of <1% and were subsequently not included in multivariable adjusted models. We assessed the relation of warfarin compared with DOAC use, and duration of warfarin use, to risk of KR or HR using conditional logistic regression in separate models, adjusting for these potential confounders.

We performed two additional sensitivity analyses. Because there may be practice variation across GP practices regarding both choice of anticoagulant and referral to surgery, we matched cases and controls according to GP practice, and adjusted for age and sex in addition to the other potential confounders listed above. Each GP practice typically serves the same geographic area but could include multiple GPs. In a second set of sensitivity analyses, we repeated the primary analysis stratified by type of joint replacement with the recognition that 97% of knee replacements are performed for knee OA, whereas hip replacements can be performed for other indications such as hip fracture.[17]

Patient and Public Involvement

Patients and the public were not involved in this study.

RESULTS

We identified 857 cases with KR or HR and matched them to 3428 controls. The mean age of both groups was 75 years and 47% were female. Other baseline characteristics are listed in Table 1. Notable differences in comorbidities included a higher prevalence of diabetes, heart failure, stroke and IHD among controls, likely because individuals with these comorbidities were less likely to be surgical candidates. As expected, obesity (BMI ≥30) was more commonly seen among cases (42.6% vs 31.7%).

TABLE 1:

CHARACTERISTICS OF PARTICIPANTS

Participants [N(%) or mean +/− SD as appropriate]	Cases (KR/HR) n = 857	Controls n = 3428
General Demographics
Mean Age, years	75.4 +− 7.2	75.4 +− 7.2
Female	403 (47.0%)	1612 (47.0%)
BMI 25-<30	313 (36.5%)	1297 (37.8%)
BMI ≥30	365 (42.6%)	1088 (31.7%)
Comorbidities
Cancer	154 (18.0%)	648 (18.9%)
COPD	214 (25.0%)	825 (24.1%)
Dementia/Cognitive Impairment	6 (0.7%)	106 (3.1%)
Diabetes	170 (19.8%)	782 (22.8%)
Heart Failure	113 (13.2%)	600 (17.5%)
Hyperlipidemia	153 (17.9%)	702 (20.5%)
Hypertension	590 (68.8%)	2303 (67.2%)
IHD	191 (22.3%)	959 (28.0%)
Mitral stenosis	8 (0.9%)	17 (0.5%)
Prior falls	153 (17.9%)	576 (16.8%)
Prior GI Bleeding	26 (3.0%)	98 (2.9%)
Prior Intracranial Hemorrhage	4 (0.5%)	37 (1.1%)
Prosthetic valve	3 (0.4%)	6 (0.2%)
Renal Disease (CKD 1–3)	203 (23.7%)	895 (26.1%)
Renal Disease (CKD 4–5 & renal transplant)	10 (1.2%)	54 (1.6%)
Severe Liver Disease	5 (0.6%)	17 (0.5%)
Stroke	135 (15.8%)	702 (20.5%)
Venous Thromboembolism	36 (4.2%)	154 (4.5%)
Medication use
Anti-hypertensives	775 (90.4%)	3076 (89.7%)
Insulin	13 (1.5%)	110 (3.2%)
Lipid lowering medications	492 (57.4%)	2090 (61.0%)
NSAIDs	337 (39.3%)	1275 (37.2%)
Oral hypoglycemic medications	87 (10.2%)	463 (13.5%)
Paracetamol	581 (67.8%)	1340 (39.1%)
GP visits (assessed within 1 year prior to 1 st warfarin/DOAC prescription)
0–5	330 (38.5%)	1606 (46.8%)
>5	527 (61.5%)	1822 (53.1%)
Hospitalizations (assessed within 1 year prior to 1 st warfarin/DOAC prescription)
0–2	820 (95.6%)	3189 (93.0%)
≥3	37 (4.3%)	239 (7.0%)

KR: knee replacement, HR: hip replacement, BMI: body mass index, CKD: chronic kidney disease, GI: gastrointestinal, AF: atrial fibrillation, COPD: chronic obstructive pulmonary disease, IHD: ischemic heart disease, NSAIDs: non-steroidal anti-inflammatory drugs, GP: general practitioner

Of the 857 cases, 64.6% were warfarin users and the remaining 35.4% were DOAC users. Among the 3428 controls, 56.1% were warfarin users and 43.9% were DOAC users. Warfarin use was associated with 59% higher risk of having a KR or HR than DOAC use (adjusted odds ratio (OR) 1.59, 95% Confidence Interval (CI) 1.31–1.92) (Table 2a).

TABLE 2:

WARFARIN USE AND RISK OF KNEE AND HIP REPLACEMENTS

a. Warfarin vs. DOAC use within one year of index date, regardless of duration:
	Cases (KR / HR)	Controls
Participants (n)	857	3428
Warfarin Use	554 (64.6%)	1923 (56.1%)
DOAC Use	303 (35.4%)	1505 (43.9%)
Odds Ratio (95% CI), matched by age and gender	1.57 (1.32, 1.86)
Adjusted* Odds Ratio (95% CI)	1.59 (1.31, 1.92)
b. Warfarin vs. DOAC use, matched by practice
	Cases (KR / HR)	Controls
Participants (n)	857	3422
Warfarin Use	554 (64.6%)	2077 (60.7%)
DOAC Use	303 (35.4%)	1345 (39.3%)
Odds Ratio (95% CI), matched by practice	1.25 (1.05, 1.50)
Adjusted** Odds Ratio (95% CI)	1.36 (1.11, 1.66)
c. Warfarin vs. DOAC use, stratified by anatomic location of joint replacement
	Cases (KR only)	Controls
Participants (n)	497	1988
Warfarin Use	324 (65.2%)	1139 (57.3%)
DOAC Use	173 (34.8%)	849 (42.7%)
Odds Ratio (95% CI), matched by age and gender	1.52 (1.21, 1.92)
Adjusted* Odds Ratio (95% CI)	1.58 (1.22, 2.04)
	Cases (HR only)	Controls
Participants (n)	485	1940
Warfarin Use	304 (62.7%)	1129 (58.2%)
DOAC Use	1173 (37.3%)	811 (41.8%)
Odds Ratio (95% CI), matched by age and gender	1.27 (1.01, 1.60)
Adjusted* Odds Ratio (95% CI)	1.33 (1.03, 1.72)

DOAC: direct oral anticoagulant, KR: Knee replacement, HR: hip replacement

^*Adjusted for the same variables as in Table 1 excluding age and sex, which were matching variables.

^**Adjusted for age and gender in addition to variables in Table 1.

Longer durations of warfarin were associated with higher risk of KR or HR compared with < 1 year of warfarin use (Figure 2 and Supplementary Table 1). Participants with warfarin use for ≥4 years had 86% higher risk of KR or HR compared with new warfarin users (<1 year) (95% CI 1.35–2.57).

Figure 2:

The relation of duration of warfarin use to risk of knee or hip replacement. Analyses adjusted for potential confounders in Table 1, with the exception of age and sex which were matching variables.

When analyses were repeated with matching by GP practice, the magnitude of the association was slightly diminished but remained statistically significant (adjusted OR 1.36, 95% CI (1.11–1.66)) (Table 2b). When we stratified analyses by type of joint replacement (knee or hip), the results were similar to the primary analysis for KRs, and slightly diminished for HRs. Warfarin use was associated with 58% higher risk of KR (95% CI 1.22, 2.04) and 33% higher risk of HR (95% CI (1.03–1.72)) compared with DOAC users (Table 2c).

DISCUSSION

In this population-based case-control study of older adults with atrial fibrillation, warfarin use was associated with a higher risk of knee and hip replacements, an indicator of end-stage OA, compared with DOAC use. Further, longer duration of warfarin use was associated with greater risk of joint replacement compared with shorter duration of its use.

The mechanism for this observed association of warfarin on risk of end-stage OA as assessed by joint replacement is most likely related to warfarin’s role as a vitamin K antagonist. Warfarin’s antagonism of vitamin K would be expected to recapitulate effects of insufficient vitamin K. Because vitamin K confers functionality to Gla proteins through gamma-carboxylation, insufficient vitamin K or inhibition of vitamin K’s functioning through warfarin leads to under-carboxylation of vitamin K-dependent proteins, limiting their functionality. An important vitamin K-dependent protein that has been specifically linked to abnormalities in soft tissue mineralization and OA is MGP. Genetic deficiencies of MGP in humans, known as Keutel syndrome, and in transgenic mice result in cartilage calcification, highlighting the role of MGP as an inhibitor of mineralization.[18–21] Of specific relevance to OA, MGP is primarily uncarboxylated in human OA cartilage, whereas it is primarily carboxylated (and therefore functional) in healthy cartilage.[22] Further, a genome wide association study agnostically identified coding variants of MGP to be associated with hand OA, and complementary functional studies demonstrated MGP RNA expression of the hand OA allele to be higher than that of the reference allele in human OA cartilage.[23] These findings complemented a smaller study that also identified a MGP SNP in hand OA.[24] Thus, the detrimental effects of warfarin through inhibition of vitamin K’s activities may be further exacerbated in those with genetic polymorphisms of MGP.

Our results add to the existing literature, extending insights regarding the importance of vitamin K and its dependent proteins in OA. Low levels of plasma phylloquinone, the major form of circulating vitamin K, were associated with prevalence of both radiographic hand and knee OA in the Framingham Offspring cohort, while low dietary vitamin K intake was associated with radiographic knee OA in a Japanese population-based cohort.[10, 11] Complementing those radiographic findings, two longitudinal studies also demonstrated an association of low plasma phylloquinone with incidence[12] and progression[13] of cartilage lesions on knee MRI, providing more direct support for a role of vitamin K in cartilage pathology. To more definitively evaluate the role of vitamin K in OA, a randomized controlled trial of vitamin K supplementation vs. placebo was conducted in 378 participants who were enrolled without regards to their baseline vitamin K status. There was no difference overall regarding prevalence of hand OA between the two arms.[14] However, in a post hoc analysis limited to those who were vitamin K insufficient at baseline, those in the vitamin K supplementation arm demonstrated 47% significantly less joint space narrowing than those on placebo, suggesting that those with insufficient vitamin K could derive benefit from vitamin K supplementation with regards to hand OA.[14]

In addition to vitamin K’s role in OA through Gla proteins in the bone and cartilage, vitamin K may have direct effects on inflammation, which may have relevance for OA.[25] Higher plasma phylloquinone was associated with lower inflammatory burden in two separate cohorts cross-sectionally.[26, 27] In contrast, under-carboxylated osteocalcin, a Gla protein, was not associated with inflammation.[26] Since these effects appear to be unrelated to vitamin K’s role in gamma-carboxylation, it is unlikely that warfarin would play a role in vitamin K’s effects on inflammation.[28] Thus, there may be potential additional benefit to targeting vitamin K in OA beyond warfarin alone.

We recognize that this observational study cannot provide definitive causal insights. However, it is unlikely that a randomized trial of warfarin vs. a DOAC for an OA endpoint would be performed. We addressed confounding by indication by limiting our sample to adults with atrial fibrillation as this diagnosis warrants anticoagulation, and by including an active comparator arm of DOACs, which are anticoagulants used for the same indication but do not antagonize vitamin K. Our study also has limitations. As with all observational studies, there is potential for residual confounding. We identified exposure to warfarin and DOACs through prescriptions, but these do not necessarily reflect medication adherence. Joint replacement was used as a proxy for end-stage OA. While approximately 97% of KRs are performed for knee OA, HRs can be performed for other reasons, such as hip fracture [17]. We are unable to disentangle putative effects of warfarin on bone density and osteoporotic fracture risk[29] versus end-stage OA as the reason for HR in this study. Nonetheless, in stratified analyses, warfarin use was associated with risk of KR with a similar magnitude as in the main analysis, and with HR, though with a slightly lower magnitude. Overall, our study provides support for a detrimental effect of warfarin in OA, complementing prior studies regarding the effects of vitamin K in OA, and supports the inference that warfarin’s effects are due to its role as a vitamin K antagonist.

Given the worldwide prevalence and impact of OA and lack of effective disease modifying therapies, our study in the context of the existing literature supports the need for a well-powered randomized control trial evaluating vitamin K supplementation in OA. Our study also raises the consideration of preferentially using DOACs over warfarin when appropriately indicated in persons with OA.

KEY MESSAGES:

• Vitamin K deficiency is associated with incidence and progression of OA.

• It is unclear whether vitamin K antagonism through warfarin is also detrimental to OA.

• In this study, use of warfarin, a vitamin K antagonist, was associated with greater risk of KR and HR (an indicator for end-stage knee OA) than DOAC use

• These data raise the consideration of using DOACs over warfarin when appropriately indicated in persons with OA.