Structural effect of avocado/soybean unsaponifiables on joint space loss in osteoarthritis of the hip




To evaluate the structural effect of avocado/soybean unsaponifiables (ASU) in the treatment of patients with symptomatic osteoarthritis (OA) of the hip.


Patients with regular painful primary OA of the hip (European League Against Rheumatism 1980 criteria) and a joint space still ≥ 1 mm (Kellgren grade 1 to 3, assessed by an independent observer prior to inclusion) entered a prospective, multicenter, randomized, parallel group, double-blind, placebo-controlled trial of 2 years duration. Patients had at least a 6-month history of regular pain and an algofunctional index (AFI) ≥ 4. The primary assessment criterion was a decrease of the joint space width (JSW) on plain anteroposterior radiographs of the pelvis performed in standing position, measured at the narrowest points by 2 independent readers, previously tested and selected and blinded to both the treatment and the time sequence. Secondary criteria were standard clinical outcome measurements (AFI, pain on a visual analog scale, consumption of nonsteroidal antiinflammatory drugs and patient's and investigator's global assessments).


One hundred sixty-three patients were included: 102 men and 61 women (mean age 63.2 ± 8.7 years). A total of 108 patients (72 men and 36 women; mean age 64 ± 7.9 years) were radiologically evaluable at 23.7 ± 2.6 months (ASU group; n = 55) and 23.7 ± 3.2 months (placebo group; n = 53). Overall comparison of the evolution of JSW showed no difference between the ASU and placebo groups, from 2.35 ± 0.93 to 1.87 ± 1.10 mm and from 2.5 ± 0.94 to 1.9 ± 1.33 mm, respectively (intergroup P value at end point = 0.9). When patients were divided into 2 subgroups according to the median value of the baseline JSW (2.45 mm), the joint space loss in the most severely affected subgroup of patients (baseline JSW ≤ median) was significantly greater in the placebo group than in the ASU group: from 1.69 ± 0.58 to 0.84 ± 0.77 mm (−0.86 ± 0.62 mm) and from 1.66 ± 0.42 to 1.22 ± 0.7 mm (−0.43 ± 0.51 mm), respectively (P < 0.01). The JSW decrease was identical, with no difference in ASU and placebo groups, in the less severely affected subgroup of patients (baseline JSW > median). Clinical parameters in the 2 groups did not differ significantly throughout the study.


This pilot randomized, double-blind, placebo-controlled trial failed to demonstrate a structural effect of ASU in hip OA. However, in a post-hoc analysis, ASU significantly reduced the progression of joint space loss as compared with placebo in the subgroup of patients with advanced joint space narrowing. These results suggest that ASU could have a structural effect but require confirmation in a larger placebo-controlled study in hip OA.


Current pharmacologic treatments of osteoarthritis (OA) are directed toward pain relief and reduction of functional impairment. At present, drugs for the treatment of OA are classified into 2 groups (1, 2): 1) symptom-modifying drugs, which can be distinguished according to their clinical kinetic profile (onset of response, long-lasting and possible carryover effect) as either rapid- or slow-acting, and 2) structure-modifying drugs, which can show a structural effect on the disease process (i.e., able to prevent, delay, stabilize, or even reverse the alteration of joint structure). This approach to drug treatment in OA has led different groups and individuals to propose recommendations or guidelines for the conduct of clinical trials to investigate new drugs in OA with reliable and reproducible methodology (1–6).

Avocado/soybean unsaponifiables (ASU) are composed of unsaponifiable fractions of avocado and soybean oils in a proportion of one-third to two-thirds, respectively. Preclinical studies of ASU have shown some anti-osteoarthritic properties. In vitro, ASU had an inhibitory effect on interleukin-1 (IL-1) and stimulated collagen synthesis in articular chondrocyte cultures (7). In another in vitro model, ASU prevented, in part, the deleterious action of IL-1 on synovial cells and abolished it on articular chondrocytes of rabbits (8). ASU may also inhibit the stimulating action of IL-1β on stromelysin, IL-6, IL-8, and prostaglandin E2 (PGE2) production and reduce IL-1β–stimulated collagenase synthesis (9). In addition, it was recently demonstrated by reverse transcriptase–polymerase chain reaction that ASU could have an anabolic effect, stimulating transforming growth factor β1 and plasminogen activator inhibitor-1 expression by articular chondrocytes (10). In vivo, ASU significantly prevented the occurrence of lesions of the bruised cartilage in the postcontusive model of OA in rabbits (11). Given the great number of elements involved in the composition of ASU, researchers have not yet identified the effective component(s).

Previous trials have shown a beneficial delayed and prolonged effect of ASU on clinical symptoms of OA of the knee and hip. In 2 placebo-controlled trials (12, 13), ASU significantly reduced pain on a visual analog scale (VAS) and functional impairment as assessed by the algofunctional index (AFI) (14). In addition, ASU significantly reduced use of nonsteroidal antiinflammatory drugs (NSAIDs) (12). The time to onset of efficacy appears to be 1 to 2 months.

The present randomized, placebo-controlled study was aimed at assessing the potential structural effect of ASU in osteoarthritis of the hip. It must be stressed that this study had been designed before publication of the recent recommendations on clinical trials in OA (1–3). Trials undertaken to show a possible beneficial structure-modifying effect in OA are long to perform and should take into account the following conditions: 1) the minimum duration is 2 to 3 years; 2) a comparison with placebo is mandatory (no reference drug to date); and 3) the evidence is based on radiologic assessment of the joint space, which narrows very slowly in common idiopathic OA of the hip or knee. Therefore, very few trials have been conducted to study the structural effect of systemic drugs in OA. With regard to hip OA, we found results reported in abstracts (15) and two studies, published as a full paper, reporting the possible beneficial effect of intraarticular therapy in knee OA (16, 17). We report here the results of one of the first pilot studies conducted to assess the structure-modifying effect of a systemic drug used in the treatment of OA.


Patients. Outpatients of both sexes, aged 50 to 80 years, were selected by private rheumatologists. Patients had to fulfill the European League Against Rheumatism (EULAR) clinical and radiographic criteria for defining primary OA of the hip (18); i.e., mechanical pain in loading position and joint space narrowing (JSN) associated with at least 1 of the following features: osteophyte and/or subchondral sclerosis and/or cyst on radiograph.

Hip OA had to be symptomatic and documented with a radiographic examination performed within the week prior to the study showing radiologic evidence of OA. Baseline radiographic examination was made by a plain standing anteroposterior view of the pelvis according to the usual technique of French radiologists (i.e., standard distance between the patient and the radiography source = 1 meter). Internal feet rotation of 15° ± 5° was required. Patients with a Kellgren/Lawrence (K/L) grade of IB to III (19) were eligible. They were required to have JSN of at least 1 mm compared with either the contralateral hip or the targeted hip on a previous radiograph. However, JSN had to be incomplete, with the remaining joint space width (JSW) ≥ 1 mm in order to have a valid baseline for the primary criterion. These figures account for 20% (JSW reduced by 1 mm) and 80% (remaining joint space = 1 mm) of a hip joint space of 5 mm, which is the normal average value (20). Prior to randomization, radiographic eligibility was appraised by an independent observer who verified the above-mentioned conditions and the quality of radiographs, which had to meet predefined quality criteria (21).

Additional selection criteria focused on disease activity and involved history of regular pain for at least 6 months before the study and an AFI score (14) ≥ 4 (range 0–24; 4 is the usual cutoff used to select symptomatic patients in hip OA).

Patients with serious concomitant medical illness, secondary osteoarthritis as defined by EULAR (18) (with the exception of mild dysplasia), the probable necessity for total hip replacement within the following 2 years, radiographic K/L grade IA or IV, rapidly destructive OA known as such at baseline, or slightly symptomatic OA were excluded from the study. Other reasons for exclusion were any intraarticular corticosteroid injection during the month prior to the study; or synoviorthesis during the 3-month period prior to the study; concomitant treatment by another so-called “symptomatic slow-acting” drug in OA, such as chondroitin sulfate, oxaceprol, or diacerhein; or not having stopped use of such treatment at least 6 months before entering the trial.

Study design. This study was a prospective, multicenter, randomized, double-blinded, placebo-controlled, parallel group trial of 2 years duration, conducted by 50 rheumatologists in Paris and the surrounding region. The study protocol was approved by the Ethics Review Board of the Cognacq-Jay Hospital (Paris) and previously accepted by a Scientific Committee responsible for blindly validating all case report forms (i.e., deciding whether they are in accordance with the study requirements and what decision to take for statistical analysis if not) before the analysis was performed.

Drug administration and concomitant therapies. After confirmation that they fulfilled the selection and inclusion criteria and after written informed consent had been obtained, patients were randomly assigned to 1 of the 2 treatment groups. One group received a 300-mg capsule of ASU (Piascledine 300, Pharmascience Laboratories, Courbevoie, France), and the other received a placebo capsule. Randomization had been previously performed by an independent statistical unit (ITEM, Paris) by blocks of 4; each investigator received 4 treatment units containing 2 ASU and 2 placebo units. ASU and placebo were supplied in coded, unidentifiable capsules for oral administration. Each patient took 1 capsule daily for 2 years. Patients entered the study according to the randomization order.

Use of NSAIDs, chosen from a list of 8 for which an equivalent daily dose was available (diclofenac, flurbiprofen, ibuprofen, indomethacin, ketoprofen, paroxen, piroxicam, tenoxicam) and/or analgesics was allowed after entry. The exact amount consumed was recorded by the patient using a self-report weekly diary. The investigator then calculated and recorded the intake for the given period. Corticosteroid injections (intramuscular, intraarticular, or epidural route) were to be avoided during the treatment period. Other so-called “slow-acting drugs in OA” were not allowed. All other concomitant medications for various medical diseases were allowed during the trial and had to be recorded.

Radiologic and clinical assessments. The radiologic examination involved plain standing anteroposterior radiography of the pelvis performed and collected at entry and after 2 years ± 6 months (year 2). Clinical assessments for efficacy and safety parameters were made by the same investigator for each patient at day 14 and months 1, 2, 4, 6, 9, 12, 18, and 24.

The primary efficacy criterion was the evolution of the JSW on plain anteroposterior radiographs of the pelvis performed in standing position. JSW was measured by 2 blinded, independent observers, each of whom read the total number of films. The observers were selected as follows: The intraobserver reproducibility of the 3 physicians (at that time in the Rheumatology Unit, Leopold Bellan Hospital) with the most experience in JSW measurement was evaluated in blind assays (2 readings 2 weeks apart) on 68 radiographs of patients with hip OA. Their coefficients of variation were 8%, 10%, and 12% (22). The 2 physicians who scored best (CC, ML) were selected for the trial radiography measurement. The intraclass coefficients of correlation were 0.97 (CC) and 0.95 (ML) with regard to intraobserver reproducibility, and 0.91 with regard to interobserver reproducibility. The measurers were unaware of the treatment given, and of the results of the other measurer. The radiographs were read by pairs, belonging to the same patient but blinded to the time sequence, and put on double-view boxes in horizontal position. The JSW measurement on radiographs at day 0 and year 2 (or earlier in case of withdrawal, with a radiograph taken before total hip replacement) was performed using a caliper with a stabilizing screw and a ½-millimeter graduated flat plastic ruler. The 2 sharp points of the caliper were put on subchondral bone limits (cortical surface of the acetabulum and bone contour of the femoral head) at the narrowest point of the joint space, then transferred directly onto the flat plastic ruler, allowing for a 0.25-mm notation. When the sharp point of the caliper fell between the mm and the ½-mm mark, the distance recorded was 0.25 or 0.75. The JSW was defined as the interbone distance represented by the distance between the points of the caliper.

Secondary assessment criteria were clinical parameters including 1) AFI score from baseline to month 12; 2) overall spontaneous pain scored on a 100-mm VAS (23); 3) quantitative NSAID consumption (among the preestablished list of 8 using an equivalence score to convert it into milligrams of diclofenac); 4) global assessment by both the patient (7-point verbal scale: I feel “much better,” “better,” “slightly better,” “identical,” “slightly worse,” “worse,” “much worse”) and the investigator (4-point verbal scale: worsening, stationary, mild improvement, marked improvement); 5) the number of days on sick leave; and 6) the number of patients requiring total hip replacement.

Safety was evaluated by: 1) the incidence of adverse effects in each group and their description; 2) the rate of withdrawal for adverse reaction; 3) the overall assessment by the patient and the physician.

Adherence was assessed by asking the patient at each visit: “Did you take all the study medication?” and by counting the number of capsules returned.

Statistical methods. Sample size was calculated preliminary to the study based on the primary criterion, variation in JSW. A previous study had shown a mean decrease of the JSW of 0.72 mm over 3 years with a very high variation (range 0–4.5 mm) and a standard deviation (SD) of 0.932 (22). The confidence interval (CI) in these conditions ranged from 0 mm to 2.58 mm. With: 1) an expected mean intergroup difference at year 2 of 0.5 mm ± 0.9 mm on the JSW variation in favor of the ASU group, 2) an anticipated patient withdrawal rate of 30% during the trial, 3) an alpha level of 0.05, and 4) a beta level of 0.20 providing a statistical power of 0.80, the estimation of the sample size was 110 available patients (55 per group). It was decided to recruit 160 patients to take into account the anticipated rate of premature discontinuation.

All randomized patients were described. The 2 groups of patients were compared at entry for demographic characteristics, OA history, and baseline values of outcome measures, using Student's t-test for continuous variables and the chi-square test for noncontinuous variables. The characteristics of the patients for whom the JSW variation was available at the end of the trial (i.e., having 2 radiographs) and those of the patients who prematurely withdrew from the study were compared to ensure that no bias had occurred in the study population during the trial.

The intent-to-treat (ITT) principle was used with the following characteristics: all randomized patients were taken into account, provided the primary criterion, JSW variation over time, was available for measurement. This method is now admitted in Europe (24). In case of withdrawal, the last observation value was carried forward, whatever its date (e.g., radiography performed a few weeks before early total hip replacement). The JSW value taken into account was the mean of the 2 values obtained by the 2 independent measurers for each film. The JSW variation was then translated into a mean annual joint space loss by dividing joint space loss by the number of months (duration of treatment) then multiplying by 12.

A secondary analysis was performed after these patients were divided into 2 groups according to the median value of the baseline JSW measurement (2.45 mm). ASU- and placebo-treated patients were compared in these 2 subpopulations: “advanced hip OA” (patients with JSW ≤ median value) and “moderate hip OA” (patients with a JSW > median value). Because this analysis had not been anticipated in the study protocol (post-hoc analysis), we present the results as an indicative secondary analysis.

An analysis of JSW variation was performed in the patients with no major deviation from the protocol to insure that the results did not differ from those of the ITT analysis. All patients presenting a major deviation from the protocol, as assessed by the Scientific Committee after the case review, were excluded from this analysis. Because the results did not differ from those of the ITT analysis, only the latter are presented. Based on the comparability of the 2 groups at entry, the efficacy analysis was carried out using Student's 2-tailed t test.

The clinical assessment parameters were analyzed using Student's t-test for comparison of the continuous variables, chi-square for dichotomous variables, and Wilcoxon's test for the ordinal categorical assessment parameters (overall efficacy evaluation by the patient and the investigator). All tests were 2-tailed, at the 5% level of significance, and confidence intervals were performed at the 95% level.

Analyses were made to identify potential factors predictive of variation in hip JSW. The correlation analysis was carried out using a multivariate approach with multiple linear regression in which the dependent or to-be-explained variable was JSW variation (in mm as a continuous variable), and the potential explicatives (i.e., independent variables) were baseline descriptive characteristics of the patients (age, sex, weight) and baseline description of OA (onset of OA, interval since first pain, unilateral or bilateral hip OA, level of pain on VAS, AFI score, NSAID consumption).


Patients. Of the 177 patients selected, 14 were judged not eligible according to the radiologic selection criteria. Thus, 163 patients entered the trial between September 1990 and December 1991: 102 men and 61 women, mean age 63.2 ± 8.7 years. A flow diagram of patients participating in this study is shown in Figure 1. Eighty-five patients were randomized to the ASU group and 78 to the placebo group. Characteristics of the 163 patients at entry are summarized in Table 1.

Figure 1.

Flow diagram of patients participating in this trial. R = randomized.

Table 1. Baseline demographic characteristics, hip osteoarthritis (OA) description, and symptomatic severity of 163 randomized patients
CharacteristicTreatment groupTotal (n = 163)
ASU (n = 85)Placebo (n = 78)
  1. Continuous variables are expressed as mean ± standard deviation. Noncontinuous variables are expressed by numbers and/or percentages. There were no significant differences between the ASU and placebo groups for each parameter. NSAID = nonsteroidal antiinflammatory drug; VAS = visual analog scale; ASU = avocado/soybean unsaponifiables.

Age (years)63.3 ± 8.763.0 ± 8.863.2 ± 8.7
Sex: Female/male36/4925/5361 (37%)/102 (63%)
Weight (kg)72.3 ± 14.268.5 ± 13.970.5 ± 14.1
Height (cm)165 ± 9165 ± 10165 ± 9
Professionally active28 (33%)22 (28%)50 (31%)
Current sport practice: Yes/No15/7012/6627 (17%)/136 (83%)
Current concomitant disease: Yes/No54/3150/28104 (64%)/59 (36%)
Uni/bilateral hip OA49/3655/23104/59
Surgery in contralateral hip7 (8%)4 (5%)11 (11%)
Kellgren/Lawrence grade
Onset of OA (months)46.5 ± 43.348.5 ± 64.047.5 ± 60.0
Interval since first pain (months)37.3 ± 32.138.5 ± 47.837.9 ± 40.2
Regular pain duration (months)14.7 ± 13.617.1 ± 21.513.9 ± 17.8
NSAID therapy at entry: Yes/No70/1566/12136 (83%)/27 (17%)
Lequesne's index score9.5 ± 3.29.4 ± 2.59.4 ± 2.9
Spontaneous pain on VAS (mm)49.9 ± 16.750.5 ± 16.850.2 ± 16.7

Sixty-eight patients (41%) withdrew from the study before completion of the 2 years of treatment. The reasons for withdrawal are listed in the flow diagram; the ASU and placebo groups were comparable with regard to these reasons. Twelve patients in the ASU group (14.1%) and 11 in the placebo group (14.1%) underwent total hip replacement surgery. However, for 13 of these 23 patients, a second radiograph was performed at the time of withdrawal, which allowed their inclusion in the ITT analysis for JSW variation. Therefore, 108 patients were available for the main analysis. The time span between the initial and final radiographs for these 108 patients was 23.7 ± 2.6 months (range 13–28) and 23.66 ± 3.22 months (range 11–27) in the ASU and the placebo groups, respectively. The final rate of subjects not available for the main analysis of the primary criterion was 55 (33.7%) of 163. There were no statistically significant intergroup (ASU/placebo) differences at baseline for demographic characteristics, OA history, or clinical status in the 108 patients eligible for the main criterion. Likewise, the baseline criterion of these 108 patients were comparable at all points to those of the 163 patients enrolled in the trial (data not shown).

Comparisons of baseline characteristics of completers and patients who withdrew without having a second radiograph did not show statistically significant differences. Patients who withdrew prematurely from the study were slightly but not significantly younger (P = 0.10) and heavier (P = 0.07), were less inclined to practice sports (P = 0.07), had a shorter time since their first pain (P = 0.090), and had a shorter time since regular pain had occurred (P = 0.08) compared with patients who remained in the structural efficacy analysis (data not shown).

Eighteen deviations from the protocol (12 in the ASU group, 6 in the placebo group) were considered “major” during the case review by the Scientific Committee. Two deviations involved inclusion criteria: 1 patient in the ASU group was aged 33 years, and 1 in the placebo group was aged 39 years. Ten patients received 1 unauthorized treatment, most of which were intraarticular injections of a corticosteroid in the knee or epidural space (5 in the ASU group, 2 in the placebo group); 2 patients in the ASU group and 1 in the placebo group took oral steroids during the trial. Four patients (3 ASU, 1 placebo) interrupted their treatment for 3 to 6 months during the treatment period. One patient in the ASU group was included for the right hip and clinically assessed for the left hip after month 4. All these patients were kept in the main structural efficacy analysis, because they had 2 radiographic examinations available.

Results of the main analysis: structure-modifying effect. Results were similar in the analysis of the ITT population and the “per protocol” analysis (data not shown). The results presented in this paper are those of the ITT analysis (n = 108) and appear in Table 2. Mean baseline JSW was 2.35 ± 0.93 mm in the ASU group and 2.50 ± 0.94 mm in the placebo group. At year 2, mean JSW in the ASU and placebo groups was 1.87 ± 1.10 and 1.90 ± 1.33, respectively, with no statistical intergroup difference (P = 0.90).

Table 2. Evolution of joint space width (JSW) during the trial in 108 patients
JS measurementASUPlaceboP
  • JS = joint space; ASU = avocado/soybean unsaponifiables.

  • *

    Mean ± standard deviation.

JSW at day 0 (mm)*2.35 ± 0.93 (n = 55)2.50 ± 0.94 (n = 53)0.40
Median (mm)2.452.45
JSW at final visit (mm)*1.87 ± 1.101.90 ± 1.330.90
Patients with baseline JSW ≤ median1.66 ± 0.42 (n = 30)1.69 ± 0.58 (n = 25)0.77
 JSW at final visit (mm)*1.22 ± 0.690.84 ± 0.770.05
 Difference day 0–final (mm)*−0.43 ± 0.51−0.86 ± 0.62<0.01
 Mean annual JS loss0.24 ± 0.310.47 ± 0.330.01
Patients with baseline JSW > median3.18 ± 0.62 (n = 25)3.22 ± 0.50 (n = 28)0.82
 JSW at final visit (mm)*2.65 ± 1.02.85 ± 0.950.46
 Difference day 0–final (mm)*−0.53 ± 0.89−0.37 ± 0.870.50
 Mean annual loss0.27 ± 0.450.22 ± 0.540.74

Because it was obvious that there were differences in the evolution of JSW according to its baseline value, it was decided to divide the population into 2 subgroups according to the median value of the JSW at baseline (2.45 mm) and to perform a post-hoc statistical analysis in these subgroups. In patients whose baseline JSW value was ≤ the median value, the mean time span between the baseline and the final radiographs was 23.32 months (SD 3.26, range 13–28) and 22.96 (SD 3.76, range 16–27) in the ASU group (n = 30) and the placebo group (n = 25), respectively (difference not significant). Twenty-five patients in the ASU group and 28 in the placebo group had baseline JSW values above the median. Table 2 and Figure 2 show the results of the analysis of JSW variation in these 2 subpopulations. In the subgroup of patients who initially had more severe narrowing, joint space loss between initial and final radiograph in the ASU group was half that in the placebo group (−0.43 ± 0.51 mm versus −0.86 ± 0.62 mm). This intergroup difference was statistically significant at P < 0.01. Conversely, there was no such difference in the ASU and placebo subgroups of patients whose baseline JSW was above the median value, whatever the variable considered: JSW at year 2 (P = 0.46), the difference between the initial and final radiographs (P = 0.50), or the mean annual joint space loss. In the ASU group, 22 patients had a radiologic change of JSW ≥ −0.5 mm, compared with 26 in the placebo group (P = 0.34). With regard to symptomatic effects in each of these subpopulations, NSAID use was similar throughout the study in the 2 treatment groups.

Figure 2.

Changes in joint space width (JSW) during the trial for the overall population and for the 2 subgroups according to baseline JSW (≤ 2.45 mm or > 2.45 mm, median value). Bars indicate means and standard errors. ASU = avocado/soybean unsaponifiables; ITT = intent to treat.

Analysis of clinical assessment parameters. Clinical parameters were analyzed during the first year of followup. Table 3 gives the results of this analysis on an ITT basis at day 0 and month 12. The results of the analysis performed on the “per protocol” population did not differ from these results (data not shown).

Table 3. Evolution of clinical parameters over the first year of followup (ITT analysis, n = 163)
Assessment parameterASU group n = 85Placebo group n = 78P
  • VAS = visual analog scale; NSAID = nonsteroidal antiinflammatory drug; ASU = avocado/soybean unsaponifiables.

  • *

    Last observation in the trial.

  • Percentages based on number of patients still professionally active (28 ASU, 22 placebo).

Lequesne's algofunctional index
 Day 09.5 ± 3.29.4 ± 2.50.79
 Year 16.4 ± 3.76.9 ± 4.70.57
Global pain on a VAS (mm)
 Day 049.9 ± 16.750.5 ± 16.80.82
 Year 131.8 ± 22.230.8 ± 26.70.81
NSAID use (mg diclofenac equivalent)
 Day 056.7 ± 54.853.5 ± 53.60.70
 Year 125.3 ± 36.825.3 ± 39.91.0
Patient's global assessment*
 Slightly better/better/much better31 (36%)30 (38%)
 Stationary32 (38%)27 (35%)0.81
 Slightly worse/worse/much worse22 (26%)21 (27%)
Requirement for total hip replacement12 (14%)12 (15%)
No. patients taking sick leave6 (24%)5 (23%)

As indicated in Table 3, there were no differences at month 12 between the 2 treatment groups regarding the clinical parameters: Lequesne's AFI, global pain assessment on a VAS, NSAID use, patient's global assessment, and investigator's global assessment. In the 2 treatment groups, the same percentage of patients (14%) required a total hip replacement during the study, and the same low proportion of patients required sick leave. The mean consumption of an NSAID (expressed in mg of diclofenac equivalent) during the 12-month period analyzed was similar in the 2 groups and showed the same trend toward reduction with time.

Factors associated with radiologic progression of joint space narrowing in 108 patients. Both the univariate regression analysis and the multiple linear regression analysis identified only 1 variable that related to variation of the JSW: baseline value of the JSW. The baseline value of the JSW accounted for 62% of the variance of JSW at year 2 ± 6 months. The final JSW value did not correlate with any of the other baseline descriptive parameters: age, sex, weight, duration since first symptoms, duration since first pain, unilateral or bilateral OA of the hip, level of pain, Lequesne's index score, or NSAID use.

Safety assessment. No severe treatment-related adverse reaction occurred during the trial. Thirty-nine patients in the ASU group (46%) and 39 in the placebo group (50%) reported 73 and 69 adverse events, respectively; these numbers were not significantly different (P = 0.60 for the comparison of the numbers of patients reporting at least 1 adverse event). Most of these events were unrelated to the treatment. Adverse reactions possibly related to treatment were those usually expected, with the most frequent being gastrointestinal, neurologic, general, and cutaneous. Four patients (2 in each group) withdrew from the study because of an adverse reaction: 2 cases of epigastralgia in the ASU group (1 after day 0 and 1 after day 30), and 1 case of epigastralgia and 1 case of arterial hypertension in the placebo group (after month 4 and day 60, respectively). Most of these reactions resolved spontaneously. The tolerability of treatment was rated “excellent” or “good” by 99% of patients in the ASU group and 97% of patients in the placebo group. The investigator's judgment gave the same results.


This paper reports the results of a pilot trial studying the structure-modifying (i.e., radiologic) effect of ASU in hip OA. The results for the overall population included in the study did not show any difference between ASU and placebo on the change in JSW over 2 years. However, in a post-hoc analysis, after having separated the study groups into 2 subpopulations according to the median baseline JSW (2.45 mm), we observed a significantly smaller decrease in JSW in the ASU-treated patients than in the placebo group. This led us to conclude that ASU may have a potential structure-modifying effect that further, larger prospective clinical investigations may confirm. In such future trials, the design should include a statistical analysis in the patients most severely affected radiologically at baseline (i.e., those under the threshold of the median value of baseline JSW). A baseline JSW ≤ 2 mm was reported by Dougados and colleagues as predictive of a worse radiologic evolution in OA of the hip compared with a baseline JSW > 2 mm (odds ratio [OR] 2.11, 95% CI 1.3–3.44) (25). It is noteworthy that in the present study, the placebo subgroups of patients (spontaneous course) with a baseline JSW ≤ 2.45 mm and those with a baseline JSW >2.45 mm showed an annual JSW loss of 0.47 ± 0.33 mm and 0.22 ± 0.54, respectively (i.e., OR 2.13, which is very close to that reported by Dougados and associates). These similar ratios confirm that our trial did study a common population of patients with hip OA.

Goker and colleagues (26), studying the progression of joint space narrowing in hip OA, showed that hip OA patients appeared to segregate into 2 subgroups: one slowly declining (0.09 ± 0.06 mm/year), the other rapidly declining (0.42 ± 0.24 mm/year). In the first study focusing on precisely measured joint space narrowing and aiming to determine the rate of joint space loss in hip OA, Lequesne and associates (22, 27) had already found different speeds of JSW deterioration: joint space loss over 3 years was 0.66 mm, with a wide range of values expressed by a large SD of 0.70 mm. The annual rate of joint space loss in our patients, whether considered globally (ASU 0.24 mm/year, placebo 0.30 mm/year) or divided into 2 subpopulations according to JSW baseline value, is consistent with that observed in this first survey (0.22 mm/year) (22, 27) and in studies reported recently (15, 26).

The results of our study raise many questions and call for comments. One should keep in mind that this study was designed in 1990. Knowledge of hip OA, its radiologic progression, and the methodology for measuring the anatomic course of OA has dramatically increased during the past 10 years (27). Some issues have therefore been discussed and resolved since this trial was started.

This study was the first to use a preinclusion independent observer for assessing the radiologic eligibility of patients, thereby answering the 2 following questions, as recommended by Ravaud and Dougados (28): 1) Is the patient affected by hip OA? and 2) Does the patient's JSW conform with the selection criteria: JSW > 20% and ≤ 75% of the normal JSW?

The radiologic examination had to be performed in standing position, as is now recommended by different groups and researchers (1, 2, 27, 28). Conrozier and colleagues have shown that JSW is influenced by weight bearing in osteoarthritic hips with a JSW ≤ 2.5 mm, but not in normal hips (29).

The assessment of anatomic progression focused on progression of JSW by measuring the interbone distance at the narrowest point. Today, this is the most recommended procedure for both hip and knee OA (1, 2, 27, 28, 30–33). The methodology used for reading and the measurement procedures in our study differed slightly from those proposed today. First, we used 2 readers, although a single reader selected as the best among several of those tested (for intraobserver reproducibility) is now preferred by some authors (1, 27, 28). This was debatable in the 1990s and remains so. For instance, Altman and colleagues proposed in 1987 to assess joint space narrowing by 3 different readers (30). Neither the recommendations of the Group for the Respect of Ethics and Excellence in Science (2) nor the Osteoarthritis Research Society International (OARSI) task force document on the design and conduct of clinical trials in patients with OA (1) prescribes the optimal number of readers to be used in such trials. This question remains open, as does the way to analyze data when several readers have been involved (e.g., a posteriori selection of the most accurate reader, mean of the readings, other). In our study, radiographs were read by pairs after masking the time sequence, which is the present means of proceeding and the most economic in terms of sample size requirement, as recently shown by Auleley and associates (34).

Second, we used a caliper and a mm-graduated ruler, not a 1/10-mm graduated magnifying glass. Reading with a 1/10-mm graduated magnifying glass directly laid over the radiograph, with the distance between 2 points marked by a very sharpened point of a special pencil, seems more accurate (27, 28). This method has better intraobserver reproducibility (35) and thus allows detection of changes with more sensitivity.

The demographic characteristics and baseline hip OA characteristics of our patients are similar to those of studies conducted more recently (15, 16). Few trials studying the structure-modifying effect of a systemic drug have been published. Therefore, few data are available at this moment. The dropout rate observed in this trial was not very different from those reported in previous papers having achieved statistical significance: 43% over 2 years in a study by Dieppe and colleagues (36) and 54% over 3 years in the Longitudinal Investigation of Nonsteroidal Antiinflammatory Drugs in Knee Osteoarthritis (LINK) Study (37) (both on knee OA), 47% in the 3-year study of hip OA conducted by Dougados and colleagues (15), and 34.5% in the study on knee OA conducted by Reginster and associates (16) (rate of patients who did not have their initial and final radiographs available to be compared with our 33.7% rate).

The clinical results concerning symptoms in this study were surprising. No difference on clinical parameters was observed between ASU and placebo groups, which contrasts with previous results significantly favoring ASU over placebo in 2 recent randomized, controlled trials of the effect of ASU on symptoms (12, 13). We do not have a clear explanation for this. One should perhaps consider the extensive use of NSAIDs by almost all patients at entry, which may introduce a high degree of background noise. Moreover, the trial was not designed to detect lightly symptomatic but radiologic changes.

In summary, from this pilot study, 2 categories of results may be presented: 1) The feasibility of joint space measurement as proposed previously (3, 22, 27) is confirmed as the pivotal method showing a relevant radiologic change in structural effect trials. 2) The second result should be presented only as an indication for future studies. Although it failed to demonstrate a structural effect in the overall population, ASU seemed to statistically significantly reduce progression of the narrowing of the joint space in a post-hoc analysis in the subpopulation of more severely affected patients, compared with those receiving placebo. These results require confirmation by a larger longitudinal, randomized, controlled trial, and their clinical relevance remains to be studied.


The authors are grateful to Dr. Hervé Caspard for his assistance in the statistical analysis of the data of this trial and to Mrs. Doreen Broneer for English revision of this manuscript. We also warmly thank all the rheumatologists who participated in this study.