Mean ± SD Western Ontario and McMaster Universities Osteoarthritis Index pain (A), stiffness (B), and function (C) subscale scores and total score (D), by treatment group over time. P values are shown where significant.
Case JP, Baliunas AJ, Block JA. Lack of Efficacy of Acetaminophen in Treating Symptomatic Knee OsteoarthritisA Randomized, Double-blind, Placebo-Controlled Comparison Trial With Diclofenac Sodium. Arch Intern Med. 2003;163(2):169-178. doi:10.1001/archinte.163.2.169
Copyright 2003 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2003
Recommendations state that acetaminophen should be used in preference to nonsteroidal anti-inflammatory drugs in the initial treatment of symptomatic osteoarthritis (OA) of the hip or knee, because of lesser toxicity and the pervasive belief that acetaminophen is not only effective in treating OA pain but is of equal analgesic efficacy as nonsteroidal anti-inflammatory drugs.
This was a randomized, double-blind, placebo-controlled trial of diclofenac sodium, 75 mg twice daily, vs acetaminophen, 1000 mg 4 times daily, in 82 subjects with symptomatic OA of the medial knee. Osteoarthritis was quantitated radiographically, and subjects met stringent baseline pain criteria. The primary evaluation of efficacy used the Western Ontario and McMaster Universities Osteoarthritis Index, with evaluations at screening, baseline, and 2 and 12 weeks after treatment. Intention-to-treat analysis was used.
Twenty-five subjects were randomized to diclofenac, 29 to acetaminophen, and 28 to placebo. The groups were closely matched for age, sex, body mass index, prior use of OA medications, baseline pain, and radiographic features. At 2 and 12 weeks, clinically and statistically significant (P<.001) improvements were seen in the diclofenac-treated group; however, no significant improvements were seen in the acetaminophen-treated group (P = .92 at 2 weeks and .19 at 12 weeks). Stratification of subjects according to baseline pain, prestudy OA medication, and radiographic grade showed no clear pattern of preferential response to diclofenac, and did not reveal a subset of subjects who responded to acetaminophen.
Diclofenac is effective in the symptomatic treatment of OA of the knee, but acetaminophen is not. A review of the literature reveals that there is scanty published evidence for a therapeutic effect of acetaminophen relative to placebo in patients with OA of the knee, because most published studies use active comparators (ie, nonsteroidal anti-inflammatory drugs) only. The advocacy of acetaminophen use in subjects with OA of the knee should be reconsidered pending further placebo-controlled studies.
LARGELY BASED on a single influential study by Bradley and colleagues,1,2 clinical guidelines for the pharmacological management of osteoarthritis (OA) of the hip and knee, emanating from both sides of the Atlantic Ocean, continue to emphasize the initial use of the pure analgesic, acetaminophen, over nonsteroidal anti-inflammatory drugs (NSAIDs).3- 7 In the initial study by Bradley and colleagues,1 acetaminophen, 4000 mg/d, was equivalent to ibuprofen, 1200 or 2400 mg/d; the investigators subsequently reanalyzed their data to show that the magnitude of underlying knee pain did not alter the apparently equivalent pain relief afforded by the pure analgesic (acetaminophen) relative to analgesic/anti-inflammatory (NSAID) therapy.2 Until recently, only one other randomized double-blind trial comparing acetaminophen, 2600 mg/d, with an NSAID (naproxen, 750 mg/d) had been published in the English-language literature, a study by Williams and colleagues8 that also showed analgesic equivalence.
Despite the recommended practice guidelines and the underlying evidence from the randomized clinical trials, patients seem to prefer NSAIDs to acetaminophen in the symptomatic treatment of OA based on perceived increased efficacy, according to 2 studies9,10 that examined the issue. Moreover, recent data support the presence of a significant inflammatory component in the pathophysiology of OA, indirectly challenging the notion that the anti-inflammatory effect of an NSAID provides no added analgesic benefit relative to the pure analgesia of agents such as acetaminophen.11,12
Some studies8,10 suggest that tolerance, particularly gastrointestinal, of acetaminophen is greater than that of cyclooxygenase-nonselective NSAIDs. However, the advent and increasingly widespread use of cyclooxygenase 2–specific NSAIDs, such as celecoxib and rofecoxib, which seem to have a more favorable gastrointestinal tolerability profile,13,14 renders the toxicity profile of acetaminophen less compelling.
As a component of an investigation of the effect of pain relief on dynamic loading in patients with OA of the medial knee,15 we performed a randomized, double-blind, 12-week trial of diclofenac sodium, 75 mg twice daily; acetaminophen, 1000 mg 4 times daily; and placebo in patients with symptomatic radiographically defined OA of the medial knee across a broad spectrum of disease, similar to what would be seen in clinical practice.
Subjects (aged 40-75 years) with unilaterally symptomatic idiopathic OA of the knee were drawn from the clinical patient population of the Section of Rheumatology, Rush-Presbyterian-St Luke's Medical Center, Chicago. To be considered for the study, patients had to meet radiographic and clinical enrollment criteria. The radiographic criteria consisted of the presence of radiographic OA (modified Kellgren-Lawrence grade16,17 ≥1) in addition to evidence of medial compartment involvement, as evidenced by possible or definite medial joint space narrowing or osteophytes.18 The clinical criteria included the presence of preenrollment ambulatory pain (defined as a visual analog scale score of ≥30 mm on the 100-mm scale corresponding to question 1 of the Western Ontario and McMaster Universities Osteoarthritis Index19 [WOMAC] pain section [visual analog version 3.0]); moderate pain, by a 5-point Likert scale; or increased pain (defined as an increase of ≥10 mm by the visual analog scale or ≥1 by the Likert scale) during a 2-week washout period following discontinuation of preexisting analgesic and/or anti-inflammatory OA medications. In addition, patients had to be capable of independent ambulation without the aid of a cane or walker and had to fall within 1 SD of weight for their height and age according to the Metropolitan Life Insurance Company tables.20 Exclusion criteria included prior intolerance to either of the study medications or a history of an NSAID allergy or intolerance, functional class I or IV,21 history of peptic ulcer disease or of significant other gastrointestinal disease, significant hepatic abnormality (aspartate aminotransferase, alanine aminotransferase, or alkaline phosphatase level >20% above the upper limit of normal), renal insufficiency (creatinine level >1.2 mg/dL [>106 µmol/L]), hematologic disease (hemoglobin level <9 g/dL, white blood cell count <4.0 × 103/µL, or platelet count <120 × 103/µL), presence of joint disease other than OA, presence of joint replacements in the lower extremity, use of substances that might interfere with pain perception (tranquilizers, hypnotic agents, or excessive alcohol intake), and anticoagulation therapy. Subjects were not permitted the use of nonstudy pain medications during the trial.
The study was a randomized, double-blind, placebo-controlled trial consisting of a screening and enrollment visit, when all prestudy OA medications were discontinued (week −2); a randomization visit corresponding to baseline (and followed immediately by treatment allocation) (week 0); and 2 study visits while taking the assigned study medication (weeks 2 and 12). In addition, a telephone interview was performed at week 6 to document any medical events and to encourage follow-up for the final (12-week) study visit. Clinical assessments were performed at weeks −2 (for the purpose of enrollment and screening), 0, 2, and 12. The study was approved and monitored by the Rush-Presbyterian-St Luke's Medical Center Human Investigation Committee (Institutional Review Board).
At baseline, standard anteroposterior weight-bearing x-ray films of the knees and the mechanical axis of the study extremity were obtained by standard clinical methods.22 X-ray films were interpreted in a random and blinded fashion by a rheumatologist (J.P.C.) experienced in the Kellgren-Lawrence scale,16 and assigned a grade according to the published method as modified for the knee by Felson et al,17 which weighs the presence of qualitative joint space narrowing equally with osteophytes in early (stage 1 or 2) OA.
This consisted of an interview and examination by a physician, and the supervised self-administration of the 2 disease-specific study instruments, the WOMAC19 (which is specific for OA of the hips or knees) and the Lequesne Algofunctional Index for the Knees (Lequesne index).23 These 2 validated indexes are the most widely used OA-specific instruments for outcome measurement in patients with OA and are recommended as core outcome measures by OMERACT (Outcome Measures in Rheumatoid Arthritis Clinical Trials).24 Subjects' conditions were evaluated by the WOMAC and the Lequesne index at enrollment (week −2), baseline (week 0) (corresponding to randomization to treatment arm), and weeks 2 and 12. The WOMAC consists of 24 questions grouped in 3 subscores: pain (5 questions), stiffness (2 questions), and function (17 questions). The Lequesne index consists of 3 sections: pain or discomfort (which includes stiffness) (5 questions), walking (maximum capable distance and use of a walking aid, comprising 2 questions), and function (4 questions). The WOMAC was selected prospectively as the instrument of primary analysis; a recent study25 suggests that it has a better responsiveness than the Lequesne index. In addition, subject and physician global assessment of arthritis activity was assessed at each visit by a 5-point Likert scale (ranging from "asymptomatic" to "very severe"). Assessment for the presence of ambulatory pain by a 5-point Likert scale was also an enrollment criterion, as previously described. Ancillary data included subject and physician global assessment of arthritis activity by a 5-point Likert scale.
Subjects were randomized to treatment with diclofenac sodium (Ciba-Geigy Corp, Summit, NJ), 75 mg twice daily; acetaminophen (McNeil Consumer Products Co, Fort Washington, Penn), 1000 mg (two 500-mg tablets) 4 times daily; or either or both of matching diclofenac and placebo capsules. The placebos were supplied by the respective manufacturers of the active medications and were identical in appearance to the active medications. All subjects, hence, took diclofenac or placebo, 1 tablet twice daily, and acetaminophen or placebo, 2 tablets 4 times daily (a total of 10 tablets daily). Compliance was assessed by pill count at each study visit.
By using the total WOMAC score as the primary outcome variable, it was estimated, based on pilot data, that to achieve 95% power to detect significant between-group differences with a significance level of .05 would require approximately 23 subjects in each treatment group, defining a clinically significant difference as 20% improvement in the total WOMAC scale.26 Data analysis was performed on an intention-to-treat basis, with imputation of missing data points as suggested in the WOMAC guidelines27 and by using last-observation-carried-forward methods.28 A 1-way analysis of variance (ANOVA) was used to compare differences in treatment groups at 0 (baseline), 2, and 12 weeks, and groupwise changes between weeks 0 and 2 and weeks 0 and 12. All post hoc analyses were performed assuming equal variances using the Bonferroni method. Paired-sample t testing (2 tailed) was used to compare changes in the WOMAC and Lequesne index and their subsections between weeks 0 and 2 and weeks 0 and 12.
Software used for analysis was Statistical Product and Service Solutions, version 10 (SPSS Inc, Chicago, Ill). Statistical significance was defined as P ≤ .05. Clinical significance was interpreted as an improvement of at least 20% in the study variable.26
Eighty-two subjects met the clinical and radiographic criteria at enrollment (week −2) and were randomized (week 0): 25 to diclofenac, 29 to acetaminophen, and 28 to placebo. The characteristics of the subjects at enrollment and randomization are given in Table 1. There were no significant differences between the groups in any of the indicated variables. Overall, more subjects took NSAIDs or high-dose (>325 mg/d) aspirin (55 subjects [67%]) than acetaminophen (3 subjects [4%]) or no prestudy OA medications (24 subjects [29%]); subjects taking NSAIDs or high-dose aspirin and acetaminophen (n = 3) were included in the NSAID/high-dose aspirin group in the percentile analysis.
Seven subjects withdrew from the trial between weeks 0 and 2 (3 in the diclofenac-treated group and 2 each in the acetaminophen-treated and placebo groups). Fourteen subjects withdrew from the trial between weeks 2 and 12 (2 in the diclofenac-treated group, 5 in the acetaminophen-treated group, and 7 in the placebo group). Overall, during the 12-week trial, 5 subjects withdrew from the diclofenac-treated group, 7 from the acetaminophen-treated group, and 9 from the placebo group. Adverse effects were the leading cause of withdrawal from the diclofenac-treated group (3 of 5 subjects) and inefficacy from the acetaminophen-treated group (5 of 7 subjects). The 9 withdrawals from the placebo group were due to inefficacy (n = 4), nonknee pain (n = 2), and other reasons (n = 3). None of the groupwise differences reached the level of statistical significance (P = .11, by ANOVA).
Pill counts performed at each study visit demonstrated greater than 90% compliance, and there was no difference in compliance between the 2- and 4-times-daily medications (data not shown).
Using the primary end point, the WOMAC, only the diclofenac-treated group was significantly improved at 2 and 12 weeks compared with baseline (Table 2). This was true for the 3 WOMAC subscales and the total WOMAC. At a significance level of P<.001, clinically significant improvement (≥20%) was seen at 2 weeks in the pain, stiffness, and function subscales and in the total WOMAC. Similar improvement was seen at 12 weeks in the diclofenac-treated group in the pain, stiffness, and function subscales and in the total WOMAC, although the level of significance was somewhat less for pain and stiffness than for function and the total WOMAC. Although the total Lequesne index was significantly improved in the diclofenac-treated group at 2 weeks and marginally improved at 12 weeks, there was variability in the significance level of the Lequesne index subscales. The Lequesne index pain and function subscales were statistically significantly improved at 2 weeks, but not at 12 weeks, and the Lequesne index distance-walked subscale was not statistically significantly improved at either time point. Clinically, greater than 20% improvement was seen in the Lequesne index pain and function subscales at 2 weeks only. Twenty percent clinical improvement in the total Lequesne index was not seen at either time point.
In marked contrast, in the acetaminophen-treated group, the results by WOMAC (subscales and sum) at 2 and 12 weeks were statistically and clinically insignificant. These results were indistinguishable from those for placebo. At 2 weeks, the Lequesne index function subscale improved in the acetaminophen-treated group and in the placebo group. The Lequesne index pain, distance-walked, and sum variables were never statistically or clinically significantly improved from baseline, mirroring the results in the placebo group. These results are summarized in Table 2.
The groupwise responses over time in the WOMAC pain, stiffness, and function subscales and in the total WOMAC are shown graphically in Figure 1.
There were no significant between-group changes in subject or physician global assessment of arthritis activity as analyzed by ANOVA (data not shown), with the exception of a difference in the physician global assessment at week 12 relative to week 0 that favored the acetaminophen-treated group (a mean ± SD improvement of 2.7 ± 0.5) relative to the diclofenac-treated group (a mean ± SD improvement of 2.4 ± 0.5), a difference of borderline statistical significance (P = .04). The mean ± SD change in the placebo group (2.4 ± 0.5) at 12 weeks was not significantly different from the change in the diclofenac- and acetaminophen-treated groups (P = .81 and .19, respectively). There were no between-group differences in the subject or physician global assessments at 2 weeks. To determine whether the response to acetaminophen or to diclofenac was in part dependent on the degree of underlying OA pain,2 subjects were stratified into tertiles of baseline pain according to the WOMAC (where the maximum pain section score is 500). The tertiles corresponded to 144.9 or less for the lowest pain tertile (27 subjects in the 3 treatment groups), 145.0 to 230.9 for the middle pain tertile (28 subjects), and 231.0 or greater for the highest pain tertile (27 subjects).
For the diclofenac-treated group, in the lowest pain tertile (n = 10), there was significant improvement by t test between weeks 0 and 2 in the WOMAC pain (52.7% improvement, P = .002), stiffness (44.7% improvement, P = .02), and function (40.6% improvement, P = .02) subscales and in the total WOMAC (43.4% improvement, P = .007). At 12 weeks, persistent marginally significant improvement was seen in the pain (29.4%, P = .04) and stiffness (24.1%, P = .05) subscales, but not in the function subscale (P = .08) or in the total WOMAC (P = .06). In the middle pain tertile, statistically significant improvement at 2 weeks was seen in the WOMAC stiffness (29.7% improvement, P = .03) and function (24.7% improvement, P = .02) subscales and in the total WOMAC (23.4% improvement, P = .02). At 12 weeks, statistically significant improvement was seen in the WOMAC function subscale only (26.3% improvement, P = .03). Finally, in the highest pain tertile, statistically significant improvement was seen at 2 weeks in the pain (29.4% improvement, P = .04) and stiffness (24.1% improvement, P = .05) subscales and at 12 weeks in the WOMAC pain (29.0% improvement, P = .03) and stiffness (24.4% improvement, P = .04) subscales and in the total WOMAC (18.8% improvement, P = .05). Responses at 2 weeks are shown by P value in Table 3.
In the acetaminophen-treated and placebo groups, in contrast, there were no statistically significant changes in the WOMAC pain, stiffness, or function subscales or in the total WOMAC at 2 or at 12 weeks for any of the 3 pain tertiles (data not shown).
Application of ANOVA according to pain tertile yielded results similar to the t test: diclofenac efficacy was inconsistent and not clearly related to the degree of baseline pain, shown in Table 4, at 2 weeks. Acetaminophen was never superior to placebo (data not shown). A sample size limitation may have compromised the power to detect a difference in the diclofenac-treated groups by t test and ANOVA.
Subjects were stratified according to prestudy OA medication (NSAID, n = 52; acetaminophen, n = 3; or no analgesic or anti-inflammatory prestudy OA medication, n = 24). Because of the few individuals who took acetaminophen before the study, these subjects were grouped along with the no prestudy OA medication group to form a non–NSAID-pretreated group (n = 27). By t test, there were no significant differences at 2 or 12 weeks for the non–NSAID-pretreated subjects in the diclofenac- or acetaminophen-treated groups (data not shown). In the NSAID-pretreated subgroup, in contrast, a statistically significant response was seen in treatment with diclofenac, but not with acetaminophen (shown for diclofenac at week 2 in Table 3). An ANOVA in general was similar, with a more consistent response to diclofenac relative to acetaminophen or placebo seen in the NSAID-pretreated subgroup compared with the non–NSAID-pretreated group (shown at 2 weeks in Table 4). Differences as a function of the prestudy OA medication regimen may, however, represent an artifact of sample size, there being roughly twice as many subjects in the NSAID-pretreated than the non–NSAID-pretreated groups.
Grouping by Kellgren-Lawrence grade into questionable or mild radiographic OA (grades 1 or 2 [n = 47]) and moderate or severe radiographic OA (grades 3 or 4 [n = 24]) showed significant improvement in both groupings at 2 and 12 weeks in the diclofenac-treated group only (t test; shown for 2 weeks for diclofenac in Table 3). The relative efficacy of diclofenac to acetaminophen and to placebo was reflected similarly by ANOVA at 2 weeks, but not at 12 weeks (data shown for 2 weeks in Table 4).
Finally, stratification by degree of qualitative medial joint space narrowing18 as 0 or 1 (n = 34) or 2 or 3 (n = 34) yielded results by t test at 2 weeks similar to those demonstrated in subgroup stratification by Kellgren-Lawrence grade: diclofenac was effective regardless of the degree of joint space narrowing (Table 3), and acetaminophen was conversely ineffective (data not shown). The efficacy of diclofenac was not seen, however, at 12 weeks (data not shown). The ANOVA by degree of joint space narrowing showed diclofenac superior to acetaminophen and to placebo at 2 weeks for grade 0 or 1 joint space narrowing, but not for greater degrees of qualitative joint space narrowing (Table 4). The efficacy of diclofenac relative to acetaminophen and placebo in this subanalysis was not seen at 12 weeks. Sample size limitation may explain the lack of statistically significant efficacy at 12 weeks or, alternately, waning of therapeutic efficacy with time.29
To determine the power of the study to reject the hypothesis of no difference between the acetaminophen and placebo groups, a post hoc analysis based on the data of Table 2 was performed: 20% improvement26 in the initial mean total WOMAC for the acetaminophen-treated group (963.5) represents a difference at 2 weeks of 192.7. The SD of this hypothetical difference was taken as 268.0, representing the SD of the difference in the total WOMAC for the diclofenac-treated group at 2 weeks. This SD is larger than that of the observed SD of the difference at 2 weeks in the acetaminophen-treated group (197.1) and, hence, is more conservative. For a size of 29 subjects in the acetaminophen-treated group, these estimates yield a calculated power to reject the 1-sided null hypothesis of no difference between the acetaminophen-treated and placebo groups of greater than 98%.
This study is in agreement with prior studies30- 32 in showing that diclofenac, 75 mg twice daily, effectively relieved knee pain in patients with OA at 2 and 12 weeks and resulted in improvements in stiffness and function; in contrast, acetaminophen, 1 g 4 times daily, did not differ from placebo in any of these variables. More subjects withdrew from the diclofenac-treated group because of intolerance to the medication rather than inefficacy; the opposite was true in the acetaminophen-treated group.
Consistent with an apparent patient preference for NSAID therapy relative to acetaminophen therapy in those with OA,9,10 2 recently published studies demonstrate that greater efficacy is obtained with NSAIDs than with acetaminophen. Pincus et al33 showed that diclofenac (administered in a regimen identical to that used in the present study) plus misoprostol was superior to acetaminophen in OA of the hip or knee in a randomized crossover study, at a cost of greater adverse effects. Similarly, Geba et al,34 in a randomized double-blind trial of acetaminophen and 2 cyclooxygenase 2–specific NSAIDs, celecoxib and rofecoxib, demonstrated that rofecoxib, 25 mg/d, was superior to acetaminophen and had a similar tolerability profile.
On the other hand, the 2 older studies cited earlier, by Bradley et al1 and Williams et al,8 reached the apparently opposite conclusion, namely, that NSAIDs (ibuprofen and naproxen, respectively) are of no greater efficacy in OA of the knee than is acetaminophen. Reconciling the seemingly disparate results of the 2 recently published studies and this study with the older and widely cited earlier investigations3- 7 may be approached by careful consideration of the differing underlying methods (summarized in Table 5). Systematically, these may be broadly classified as differences in (1) study design and analysis, (2) patient and disease selection, and (3) outcome measurement and interpretation.
Regarding study design and analysis, to our knowledge, this study represents only the second published, randomized, double-blind, placebo-controlled trial of acetaminophen in patients with OA (MEDLINE search, keywords osteoarthritis or osteoarthrosis and acetaminophen or paracetamol), following a study in 25 patients with OA of the knee by Amadio and Cummings35 that showed efficacy of acetaminophen at a dose of 1 g four times daily (the dose used in the present study). The studies by Bradley,1 Williams,8 and Geba34 and colleagues were randomized double-blind studies without a placebo, and the study by Pincus et al33 was a randomized, double-blind, crossover study (also without a placebo). Hence, all comparison studies of NSAIDs vs acetaminophen subsequent to that of Amadio and Cummings are grounded in the fundamental belief of the validity of that 1 small study.
Second, research practice dictates that primary analysis be undertaken with intention to treat, generally using last-observation-carried-forward methods for data imputation.36,37 The older studies either did not use8 or did not indicate the use1,35 of such methods (which had not yet been widely accepted as the standard); more recently, Pincus et al33 did not indicate the use of intention to treat/last observation carried forward and Geba et al34 used a modified intention-to-treat protocol that did not use last observation carried forward for the primary (WOMAC) data. Finally, the efficacy of pharmacological therapy in patients with OA may wane with time29; however, the studies cited were all of roughly the same medium-term duration. The present study (at 12 weeks) is longer. Nevertheless, there seemed to be some diminution in the efficacy of diclofenac at 12 weeks relative to 2 weeks (Table 2 and Figure 1).
Concerning patient and disease selection, these differed substantially in the NSAID/acetaminophen studies cited. All but the study by Geba et al34 used radiographic criteria of various degrees of rigor, and all required the presence of knee pain (or, in the case of the study by Amadio and Cummings,35 tenderness, swelling, or warmth). The studies by Geba34 and Pincus33 and colleagues and the present study required a quantitated minimum degree of knee pain. The pain requirement, however, differed in the study by Geba et al, depending on whether the prestudy OA medication was acetaminophen or an NSAID, which may have potentially affected the differential pain responses in the researchers' noncrossover design. All 5 published studies and the present one were of OA of the knee; OA of the hip was the disease in a few (22%) of the subjects in the study by Pincus et al. Most (55 [67%]) of the subjects in the present study were taking NSAIDs or high-dose aspirin before the study, which may reflect relatively severe manifestations of disease and/or the rheumatology clinic population out of which the patients were recruited.38 Of the studies in Table 5, only the study by Geba et al describes the nature of the subjects' prestudy OA medications: 77% were taking NSAIDs in that study, possibly reflecting disease more severe than that in the present study. Prior treatment with diclofenac did not abrogate an apparent clinical response on subsequent treatment with acetaminophen in the study by Pincus et al.
Last, in outcome measurement and, above all, in interpretation, there were differences between the present and prior studies. Numerous investigations25,39- 41 have demonstrated the superiority in patients with OA of disease-specific instruments such as the Lequesne index and, especially, the WOMAC. The older studies1,8,35 in Table 5 appeared before the widespread acceptance of these instruments and perforce relied on less accurate measures, such as a single-question visual analog scale or Likert scale scores. The reliance on the joint-specific targeted WOMAC (directed toward a single joint rather than toward the whole lower extremities in the conventional WOMAC application19), as used by Pincus et al,33 is unique among the recent studies, but has been used successfully in the evaluation of surgical outcome in patients with OA of the hip.42
The most compelling explanation for the disparate results of these studies (ie, whether acetaminophen is as effective as NSAIDs or, indeed, is an effective analgesic at all in patients with OA of the knee) may be the emphasis placed by the researchers on their interpretation of their results. The study by Bradley et al1 found a statistically significant pain response for acetaminophen in only 1 of 3 variables investigated, yet concluded that acetaminophen was equivalent to high-dose ibuprofen (which itself was significantly effective in all 3 variables). Similarly, Williams et al8 found acetaminophen effective in 1 of 2 variables and naproxen effective in both of them and superior to acetaminophen in 1 of them, yet concluded that efficacy was similar between the 2 drugs. The study by Pincus et al33 showed that diclofenac was more effective than acetaminophen, but does not indicate the presence or lack of statistical or clinical efficacy for either treatment arm alone and, like the study by Geba et al34 (and the studies by Bradley1 and Williams8 and colleagues), lacks a placebo control.
The relatively few patients examined in this study limited the power of subgroup analysis (Table 3 and Table 4), specifically, the capacity to clearly discern whether some clinical subsets of patients with OA of the knee derive greater benefit than do others, as suggested by the Pincus et al,33 but not the Bradley et al,1,2 data. On the other hand, this study, although smaller than all but the 1 investigation that showed a statistically significant effect of acetaminophen vs placebo (by Amadio and Cummings35), used the rigorously validated, currently recommended outcome assessments19,23,36; is placebo controlled; and is of sufficient power (>98%) to argue compellingly for the lack of substantial efficacy of acetaminophen in subjects with clinically and radiographically quantified idiopathic OA of the knee. Thus, whereas the bulk of the literature that ascribes substantial benefit to acetaminophen in patients with OA is based on trials of active comparators, it seems from this study that the effect vs placebo may be considerably more modest, and in the absence of larger scale, randomized, placebo-controlled trials of acetaminophen in patients with OA of the knee, the widespread emphasis5- 7 on its use in patients with OA should be reconsidered.
Corresponding author and reprints: John P. Case, MD, Section of Rheumatology, Rush Medical College, Rush University, 1725 W Harrison St, Suite 1017, Chicago, IL 60612 (e-mail: firstname.lastname@example.org).
Accepted for publication June 11, 2002.
This study was supported in part by a Specialized Center of Research osteoarthritis grant AR39239 from the National Institutes of Health, Bethesda, Md; and an intramural development grant from the Rush Arthritis and Orthopedics Institute, Chicago, Ill (Dr Case). Mr Baliunas was the recipient of a Dean's Summer Research Fellowship from Rush Medical College, Chicago.
We thank Richard F. Loeser, MD, for a critical reading of the manuscript.