Do NSAIDs Affect Longitudinal Changes in Knee Cartilage Volume and Knee Cartilage Defects in Older Adults?

Materials and Methods 

Results 

A total of 1100 subjects (51% female) aged between 51 and 81 years (mean 62 years) participated in the Tasmania Older Adults Cohort study. Of those, 122 had no knee MRI scans at baseline due to the following reasons: contraindication to MRI (33 claustrophobia, 8 metal in the body, and 2 pacemaker), 5 had joint replacement, and the rest refused or were unable. Over 2.9 years, 149 subjects were lost to follow-up study due to: 28 deceased, 20 moved to other states or overseas, 15 had joint replacement, 28 physically unable, and others refused or no reason. The remaining 829 subjects (85% of those originally studied with MRI scans) who completed the study agreed to have follow-up. The first 419 had the second MRI scans but not the others as the MRI machine in the hospital was updated and was not available for research purposes. Twenty-four participants who had both types of NSAIDs were excluded from analyses. There were no significant differences between these subjects and the rest of the sample (Table 1). Fifty-five percent of subjects had radiographic changes (joint-space narrowing or osteophytes score of ≥1) in the right knee and 52% had knee pain.

Table 1. Comparisons at Baseline Between the Current Sample and Subjects Without Follow-up Magnetic Resonance Imaging

		Current Sample (n=395)	Others (n=583)	P Value
	Age (years)	62.6 (7.3)	62.2 (7.5)	.41
	Female (%)	50	50	.87
	Body mass index (kg/m2)	27.6 (4.5)	27.8 (4.8)	.57
	Radiographic osteoarthritis (%)	57	61	.21
	Total cartilage volume (ml)	8.3 (1.9)	8.1 (2.0)	.59
	Total cartilage defects	5.8 (2.3)	5.8 (2.5)	.98
	WOMAC pain	8.4 (6.4)	8.7 (5.9)	.44
	Users of COX-2 inhibitors (%)	11.7	9.0	.17
	Celecoxib, 100-200 mg (%)	7.9	5.1	.08
	Rofecoxib, 12.5-25 mg (%)	3.8	3.9	.95
	Users of conventional NSAIDs (%)	5.7	6.6	.59
	Diclofenac, 50-100 mg (%)	3.3	4.2	.46
	Ibuprofen, 400 mg (%)	0.8	0.8	.99
	Naproxen, 500-1000 mg (%)	0.5	0.7	.65
	Ketoprofen, 200 mg (%)	0.2	0.3	.75
	Piroxicam, 20 mg (%)	1.0	0.5	.44

WOMAC=Western Ontario McMaster Osteoarthritis Index; COX-2=cyclooxygenase-2; NSAIDs=nonsteroidal anti-inflammatory drugs.

Characteristics of the subjects are presented in Table 2. Users of NSAIDs (including conventional NSAIDs and selective COX-2 inhibitors) and nonusers of NSAIDs were similar in terms of sex proportion, smoking status, lower-limb muscle strength, knee cartilage volume, and bone marrow lesions. However, compared with nonusers of NSAIDs, users of selective COX-2 inhibitors were older and had higher BMI; higher prevalence of rheumatoid arthritis, knee pain, and osteophytes; greater WOMAC knee pain score; greater medial tibial bone area; and higher prevalence of cartilage defects in the lateral compartment; users of conventional NSAIDs had higher prevalence of rheumatoid arthritis and osteophytes. Users of conventional NSAIDs also had more prevalent knee pain (P=.055) and joint-space narrowing score in the medial compartment (0.96±0.86 vs 0.61±0.70, P=.02).

Table 2. Characteristics of Participants

		COX-2 Inhibitors Users (n=40)	Nonusers (n=334)	NSAIDs Users (n=21)
	Age (years)	65.7(7.6)⁎⁎	62.2(7.2)	61.3(5.6)
	Female sex (%)	47	50	58
	Body mass index (kg/m2)	29.4(4.2)⁎⁎	27.4(4.5)	27.8(4.0)
	Smokers (%)	37	50	50
	Steps per day	8663(3820)⁎	9823(3653)	9449(3208)
	Lower-limb muscle strength (kg)	90.9(45.0)	95.0(49.3)	89.3(47.7)
	Rheumatoid arthritis (%)	23⁎⁎	8	25⁎⁎
	Knee pain (%)	69⁎⁎	45	65
	Knee pain WOMAC score (0-45)	5.9(8.0)⁎⁎	2.9(6.1)	5.2(5.2)
	Right knee joint space narrowing (%)	67	54	71
	Right knee osteophytes (%)	15⁎	6	25⁎⁎
	Prevalent medial cartilage defects (%)	22	21	29
	Prevalent lateral cartilage defects (%)	33⁎	19	17
	Prevalent patellar cartilage defects (%)	47	37	38
	Medial tibial cartilage volume (mL)	2.4(0.7)	2.3(0.6)	2.2(0.6)
	Lateral tibial cartilage volume (mL)	2.7(0.6)	2.8(0.7)	2.6(0.9)
	Patellar cartilage volume (mL)	3.3(0.9)	3.4(0.9)	3.4(0.8)
	Medial tibial bone area (cm2)	22.0(3.2)⁎	21.1(3.0)	20.9(3.7)
	Lateral tibial bone area (cm2)	12.5(2.2)	12.2(2.1)	12.1(2.2)
	Medial bone marrow lesions (0-6)	0.6(1.1)	0.3(0.8)	0.6(0.9)
	Lateral bone marrow lesions (0-6)	0.5(0.9)	0.3(0.8)	0.4(1.0)

COX-2=cyclooxygenase-2; NSAIDs=nonsteroidal anti-inflammatory drugs (conventional); WOMAC=Western Ontario and McMaster Universities index. Mean (SD) except for percentages.

⁎ P <.05. ⁎⁎ P <.01 vs nonusers of NSAIDs and COX-2 inhibitors.

In unadjusted analysis, over 2.9 years, users of conventional NSAIDs had greater increases in cartilage defects in the medial (54% vs 33%, P <.05) and lateral (54% vs 30%, P <.05) tibiofemoral compartments than nonusers of any NSAID, but users of selective COX-2 inhibitors had no greater increases in cartilage defects. After adjustment for all potential confounders, users of conventional NSAIDs had 2.6- to 3.1-fold greater odds of increases in cartilage defects in the medial and lateral tibiofemoral compartments than nonusers of NSAIDs, whereas users of COX-2 inhibitors had a 60% reduced odds of an increase in cartilage defects in the medial compartment (Table 3). Compared with users of COX-2 inhibitors, users of conventional NSAIDs had higher but nonsignificant odds of an increase in cartilage defects in the medial (odds ratio 2.2, 95% CI, 0.5-11.1) and lateral (odds ratio 2.2, 95% CI, 0.5-9.6) tibiofemoral compartments.

Table 3. Associations between Use of NSAIDs, Increases in Knee Cartilage Defect, and Loss of Cartilage Volume Over 2.9 Years

		Increases in Defects⁎ OR (95% CI)	Cartilage Loss pa⁎ β (95% CI)
	Use of COX-2 inhibitors vs nonuse
	Medial tibiofemoral	0.4(0.2-1.0)†	–0.03(–1.8-1.8)
	Lateral tibiofemoral	1.3(0.6-2.9)	0.9(–0.8-2.6)
	Patellar	0.8(0.3-2.1)	0.1(–1.9-2.0)
	Use of conventional NSAIDs vs nonuse
	Medial tibiofemoral	3.1(1.0-9.1)†	–1.1(–3.4-1.2)
	Lateral tibiofemoral	2.6(1.0-6.7)†	–1.1(–3.1-0.8)
	Patellar	0.2(0.1-1.1)	–1.0(–3.3-1.3)

COX-2=cyclooxygenase-2; NSAIDs=nonsteroidal anti-inflammatory drugs; OR=odds ratio; CI=confidence interval. Dependent variable: Increase in cartilage defects or loss of cartilage volume per annum (pa). Independent variable: use of COX-2 inhibitors (or use of conventional NSAIDs) vs nonuse of COX-2 inhibitors and conventional NSAIDs.

⁎ Adjusted for sex, age, body mass index, knee pain, smoking, lower-limb muscle strength, baseline cartilage volume (or cartilage defects), tibial bone area, subchondral bone marrow edema, radiographic osteoarthritis, and rheumatoid arthritis. † Statistically significant result.

In unadjusted analysis, over 2.9 years, users of conventional NSAIDs had greater loss of lateral tibial cartilage volume than nonusers of NSAIDs (−3.9% vs −1.8%, P <.05), and users of selective COX-2 inhibitors had no greater loss of cartilage volume. After adjustment for potential confounders, users of selective COX-2 inhibitors had no greater loss of cartilage volume, whereas users of conventional NSAIDs had greater loss of cartilage volume per year, although these were not statistically significant (Table 3). Compared with users of selective COX-2 inhibitors, users of conventional NSAIDs had greater loss of lateral (−3.9% vs −1.7%, P <.05) but not medial tibial cartilage volume in unadjusted analyses, but had greater loss of both tibial cartilage volume per year (medial: −5.3% vs −3.1%; and lateral: −3.6% vs −1.1%) in multivariable analyses (Figure).

View Large Image Download to PowerPoint

Figure. Comparison of tibial cartilage loss per annum over 2.9 years between use of selective COX-2 inhibitors and use of conventional NSAIDs. Cartilage loss was adjusted for sex, age, body mass index, knee pain, smoking, lower-limb muscle strength, baseline cartilage volume, tibial bone area, subchondral bone marrow edema, radiographic osteoarthritis, and rheumatoid arthritis.

Discussion 

This study is the first utilizing MRI to examine the possible effects of NSAIDs use on knee structural changes in older people. Compared with nonusers of NSAIDs, users of conventional NSAIDs had about 3-fold greater odds of an increase in cartilage defects in the tibiofemoral compartments, whereas users of COX-2 inhibitors had 60% reduced odds of an increase in cartilage defects in the medial compartment over 2.9 years. Furthermore, we found that users of conventional NSAIDs had higher loss of tibial cartilage volume than the users of selective COX-2 inhibitors over time. These associations persisted after adjusting for many potential confounders including knee pain, radiographic osteoarthritis, and rheumatoid arthritis.

NSAIDs are very widely used, so it is important to understand their effects on osteoarthritis disease progression. Some studies in the 1990s suggested that conventional NSAIDs such as naproxen, ibuprofen, and diclofenac could inhibit the synthesis of cartilage proteoglycans and increase the release of proteoglycans in vitro.20, 21 In a rat experimental model of osteoarthritis, both diclofenac and indomethacin were found to have deleterious effects on articular cartilage, and caused a statistically significant increase in the severity of most osteoarthritic parameters.6 In clinical studies, Dieppe et al22 and Williams et al23 reported that there was no significant difference in progressive joint space narrowing between diclofenac- or naproxen-treated patients and paracetamol- or placebo-treated patients over 2 years; however, Reijmans et al10 reported that the chronic use of diclofenac accelerated progression of knee and hip osteoarthritis in a population-based prospective cohort study. The other NSAIDs (ibuprofen, naproxen, or piroxicam) were not associated with progression.

The clinical trials22, 23 performed so far had high dropout rates, and the rate of radiographic progression in the withdrawals is unknown. In addition, conventional radiography has limited sensitivity to change, is less sensitive to focal lesions than to general cartilage loss, and cannot demonstrate the pattern of cartilage and other structural changes throughout the joint surface. Therefore, measurement of the joint space width in radiographs is not ideal for reliable evaluation of cartilage loss and surface alterations. In the population-based study,10 the results could be influenced by many confounding factors such as baseline joint pain.24 Therefore, these previous clinical studies are somewhat inconclusive.

Assessment of cartilage defects using MRI is highly reproducible,17 and correlated with histological25 and arthroscopic26 finding. MRI assessment is far more sensitive than radiographic assessment. In a relatively healthy cohort, the prevalence of radiographic osteoarthritis was 18% but the prevalence of cartilage defects was 57%.27 Knee cartilage defects are associated with knee pain in 2 separate cohorts, are predictive of cartilage loss in multiple studies, and also joint replacement.27 Knee cartilage defects have a highly variable natural history, with improvement being reported in a number of groups worldwide. This improvement can be predicted by young age, lower BMI, decrease in BMI, small tibial bone size, and no radiographic changes.27

Using MRI assessments on loss of knee cartilage volume and increases of knee cartilage defects, we found that, in this population-based prospective study, uses of conventional NSAIDs (including diclofenac, ibuprofen, naproxen, ketoprofen, and piroxican) had significantly higher risks of increases in knee cartilage defects, and greater but nonsignificant loss of tibial cartilage than nonusers of NSAIDs over 2.9 years. All these associations were fully adjusted for possible confounding factors, including knee pain, radiographic osteoarthritis, age, BMI, female sex, smoking, steps per day, lower-limb muscle strength, tibial bone area, cartilage defects, cartilage volume, subchondral bone marrow edema, and other disease status such as rheumatoid arthritis. This finding is consistent with the previous report10 suggesting that conventional NSAIDs have an adverse effect on cartilage, which is independent of confounding factors and is most likely due to direct inhibition of constitutively expressed COX-1.7

Selective COX-2 inhibitors may have beneficial effects on cartilage in vitro, but in vivo studies have not confirmed this. Celecoxib showed a significant beneficial effect in vitro, not only in late-stage osteoarthritis but also in more early-stage osteoarthritis,28 and was able to increase proteoglycan synthesis and diminish proteoglycan release in osteoarthritis cartilage obtained from joint replacement surgery.7, 9 In a canine groove model of osteoarthritis, celecoxib treatment did not alter any characteristics of early osteoarthritis, suggesting a chondroneutral effect of celecoxib in vivo.8 In a 12-month, multicenter, prospective and open-label trial,29 celecoxib (200 mg/day to 400 mg/day) was shown not to accelerate progression of osteoarthritis of the knee over a 1-year period; however, this was an uncontrolled study. Recently, Sawitzke et al30 reported in the Glucosamine/Chondroitin Arthritis Intervention Trial that celecoxib 200-mg daily treatment of knee osteoarthritis for 24 weeks did not affect joint space width compared with the placebo; however, it showed a trend toward improvement relative to the placebo in patients with a Kellgren/Lawrence score of 2. However, radiograph-based studies need to be much longer than 24 weeks to provide robust results. The current clinical study is the first to demonstrate that selective COX-2 inhibitors (celecoxib and rofecoxib) can significantly reduce knee cartilage defect progression or development, suggesting that they might have chondroprotective properties. Furthermore, we found that users of selective COX-2 inhibitors had less loss of cartilage volume than users of conventional NSAIDs over 2.9 years, suggesting that selective COX-2 inhibitors are superior to conventional NSAIDs in the treatment of osteoarthritis in terms of slowing osteoarthritis progression.

This is an observational study, and channeling bias caused by the imbalance of baseline disease and participants' characteristics between groups could be a concern. Knee pain, female sex, increasing age, and radiographic osteoarthritis were associated with increased loss of cartilage volume and increases in cartilage defects in this sample (data not shown), so the higher prevalence (or proportion) of these factors in the users of COX-2 inhibitors would be expected to be associated with greater cartilage loss and increased cartilage defects. However, the actual results were opposite to this (with COX-2 inhibitor users having less cartilage loss and decreased cartilage defects), which is not consistent with channeling bias. Furthermore, the multivarable model led to some strengthening of the differences between NSAIDs and COX-2 inhibitors, indicating that some of these factors were acting as weak negative confounders (not positive confounders), again not providing support for channeling bias. This model also is fully adjusted for these factors (to the extent that they can be measured accurately), thus a correct interpretation is that the association appears independent of these factors, as the model demonstrates the effect is present in a subject with the same age, BMI, sex, pain, and radiographic osteoarthritis. This does not negate the need for a well-designed randomized controlled trial to test this hypothesis, but does suggest a biologic effect.

Conclusions 

Nonselective NSAIDs may have deleterious effects, while selective COX-2 inhibitors may have beneficial effects on knee cartilage. Randomized controlled trials examining knee structure to confirm this finding are warranted.