1.2. Methods

1.2.1. Trial design

The design of this study was a post-hoc analysis of our multicentre 1:1 randomised controlled trial focused on patients with knee OA (www.clinicaltrial.gov: NCT00348777) [11], which was conducted in accordance with the Helsinki declaration 2000. The original study was carried out in the three largest spa therapy resorts in France, Aix-les-Bains, Balaruc-les-Bains and Dax. In total 29,000, 36,000 and 50,000 people, respectively, attended these three resorts for spa therapy in 2007.

1.2.2. Participants

Patients were recruited between June 2006 and April 2007 from the local areas surrounding the spa centers thus facilitating attendance to treatment sessions on a daily basis. Recruitment was carried out using advertisements in the regional press and posters in pharmacies and general practitioners’ waiting rooms [9]. Advertisements referred to treatment for knee OA but did not specify spa therapy or any other specific treatment modality. Patients were assessed for suitability and enrolled into the study by three trained physicians in private practice outside and independent of the spa settings and had no vested interest in the spa centers. Participants were included in the original study if they had knee OA confirmed by physical examination and the presence of osteophytes on the X-rays [21]. Exclusion criteria were:

•

knee OA limited to the patellofemoral joint;

•

severe depression or psychosis;

•

a contraindication (immune deficiency, evolving cardiovascular conditions, cancer, infection) or intolerance to any aspect of spa treatment;

•

professional involvement with a spa resort;

•

spa treatment within the previous 6 months;

•

intra-joint corticosteroid injection to the knee within the past 3 months;

•

massage, physiotherapy or acupuncture in the past month; a non-steroidal anti-inflammatory drug (NSAID) within the past 5 days or other analgesic drug in the previous 12 h, or a change in symptomatic slow-acting drugs prescribed for OA (SYSADOA) in the past 3 months.

For this post-hoc study the participants were retrospectively classified into two groups (GOA and no proof of GOA) according to the following inclusion criteria: pain > 30 mm on visual analogue scale (VAS) with knee OA (ACR criteria [1]) and a diagnosis of GOA fulfilling the Kellgren criteria [22] (hand OA) and/or the Dougados criteria [10] (hand OA or bilateral knee and spine OA) and/or the ACR criteria [1] (spine OA and two OA localisations). For hand or foot OA, a clinical diagnosis of OA was accepted. For the diagnosis of OA in other localisations e.g. spine, hip, ankle, and wrist, an X-ray examination indicating the presence of osteophytes was required. For this part of the study, diagnostic criteria data for GOA was collected in the “prognostic factors” section of the electronic case report form completed by the physician at initial assessment. In some cases, when the side of the knee OA was not reported by the examining physician, the diagnosis was considered to be unilateral knee OA. Additionally, the localisation of the joint or the level of spine affected by OA had to have been reported.

1.2.3. Randomisation

For all the knee OA patients the randomisation technique of Zelen was used [37] and [38]. The Zelen randomisation method [37] and [38] implies that patients are not informed of the existence of two different treatment groups. If the patient refused to participate as randomised, he was proposed the other treatment, but remained in their assigned group for intention-to-treat analysis.

Eligible patients were randomly assigned to spa therapy or to the control group using a centralised computer programme. Randomisation was stratified by centre and in blocks of 8 with random order. For this post-hoc analysis the patients were kept in their original randomly assigned group. The three examining physicians were trained in the randomisation technique and in how to complete the electronic case report forms (Clininfo, Lyon) under their normal consultation conditions by a clinical research assistant from the coordinating centre, the clinical research centre, Grenoble, France.

Physicians reported the identification of the patients, the inclusion and exclusion criteria before randomisation into the treatment or control groups. Retraining of the examining physicians in the randomisation method was planned if over 15% of patients had changed groups, and the randomisation technique would have been altered if over 30% changed.

1.2.4. Blinding

In order to conceal the existence of the other group and realise a partial blinding of patient [37] and [38], randomisation was performed before written informed consent was obtained. Patients were told only about the group to which they were assigned and given one of the two possible patient information documents with the consent form. In addition, delocalisation of the consultation away from the spa setting was done with the intention of keeping patients ignorant of the other group. Researcher concealment from allocation was assumed by a protected computer file on a website developed for this purpose and used for the randomisation process.

Spa employees were concealed to allocation and partially blinded to the treatment group. Study patients were mixed in with the general population of patients treated in the centre, and the centre personnel were not informed which patients were taking part in the clinical trial.

1.2.5. Interventions

At enrolment, the examining physician orally explained the standardised home exercise programme (HEP) as proposed by O’Reilly [29] and evaluated by Ravaud [30] (see supplementary material), to all patients. The importance of performing six repetitions of each of the four exercises, three times daily, was emphasised. All patients from both groups were given a booklet about knee OA including details of the HEP and continued their usual treatments (analgesics, NSAIDS, SYSADOAs, and/or physiotherapy).

Those participants randomised into the spa therapy group received 18 days of therapy over three weeks. The treatments were detailed in a prescription issued by an independent physicians (GP, rheumatologist or physiatrist) attached to the centre (spa physician). Spa mineral water and treatments are approved and controlled by the French authorities. The spa intervention consisted of the following treatments:

•

general shower of 3 minutes at 38°;

•

mineral hydro-jet sessions at 37 °C for 15 minutes;

•

manual massages of the knee and thigh under mineral water at 38 °C by a physiotherapist for ten minutes;

•

applications of mineral-matured mud at 45 °C for fifteen minutes;

•

supervised general mobilisation in a collective mineral water pool at 32 °C in groups of six patients for 25 minutes.

The supervised general mobilisation was delivered by a physiotherapist who had been trained in the protocol. This supervised mobilisation programme is not specific to knee OA and also mobilised the spine, shoulders, wrists, hips and ankles. The duration of each treatment was timed by stopwatch and water temperature controlled automatically. The chemical composition of the spa treatment products in each centre was constant. Attendance, tolerance and proper performance of the various treatments was checked during the three visits with the spa physician, at the beginning, in the middle and at the end of the 3-week therapy period. In all centres the physiotherapists and hydrotherapists involved in the study were experienced and state registered. Study patients were treated in exactly the same way as non-study patients. The attendance of the patient to all the treatment session was verified by the executives of each spa centre.

Patients in the control group were offered a three-day wellness package at the spa resort following the six-month follow-up visit.

1.2.6. Outcomes

Impact of knee OA on functioning was measured using the Western Ontario and McMasters Osteoarthritis Index (WOMAC) [3]. A five point Likert scale version of the WOMAC was used for each item, with higher scores indicating greater disability. Pain was measured using the visual analogue scale (VAS) [20]. Quality of life was measured using the validated French version of MOS SF36 (SF36) [26]. The final scores were divided in to two summary measures, the physical component summary score (PCS) and the mental component summary score (MCS). The self-assessments forms were completed by the patients without assistance in the waiting room. Knee flexion, effusion and swelling; associated treatments; patients’ and examining physicians’ overall opinions of outcome (worse, neither worse nor better, better) and adverse events were collected by the same external physicians in each centre who conducted the assessments at baseline and during follow-up visits at 3 and 6 months. At 9 months follow-up the self-assessment forms were completed at home.

The primary outcome for this study was the number of patients achieving Minimal Clinically Important Improvement (MCII) at the sixth month follow-up and no knee surgery. MCII is defined as ≥ −19.9 mm on the VAS pain scale and/or ≥ −9.1 points in a WOMAC function subscale [33]. Secondary outcomes were: MCII at 3 and 9 months, “patient acceptable symptom state” (PASS) defined as VAS pain ≤ 32.3 mm and/or WOMAC function subscale ≤ 31 points [34] and significant improvement in WOMAC and quality of life (SF36) [26].

1.2.7. Sample size

Using an open preliminary study on 13 usual consecutive patients with knee OA we calculated that 50% could be improved in the spa group and an estimated 25% could improve in the control group. Thus, if the agreed alpha risk is 5% and the beta risk 20%, 58 patients per group or 67 allowing for 15% loss to follow-up was required for this analysis.

1.2.8. Statistical methods

The analysis for the subgroup patients involved an unplanned post-hoc analysis of data from our previous randomised controlled trial [11]. The analysis was based on intention-to-treat principles. We used numbers and frequencies for qualitative variables and averages and standard deviations for continuous variables.

All relevant clinical and demographic variables were tested according to the two a posteriori defined groups. Given that randomisation, considered non-stratified for GOA, cannot guarantee the balance of prognostic factors between the two groups a Chi² test was applied once the validity conditions were verified. Otherwise a Fisher's exact test was applied for qualitative variables, and Student's t-test was used for continuous variables. Statistical tests were performed with the type I error rate set at α = 0.05.

The primary criterion was tested using a Chi² test without continuity correction. We also reported the crude odds ratio (OR) and the corresponding 95% confidence interval (CI). Additionally, a multivariate step-down logistic regression to estimate the adjusted OR and the corresponding 95% CI associated with the group to explain the result at six months was performed. A random-effects model was used to account for between-centre differences. The explanatory variables entered into the model were the VAS pain score, the WOMAC pain score, and the WOMAC baseline function score.

Secondary qualitative criteria was also tested by the Chi² test without continuity correction: the mean six-month differences in the VAS pain score and the WOMAC function score between the two groups are compared using Student's t-test. We also calculated the “effect size”. An ANOVA (analysis of variance) for repeated measures was performed to study the physical and psychological dimensions of the SF36 datasheets at baseline and at six months.

The analyses were performed using the STATA software package (version 10.0; Stata Corp, College Station, Texas). The coordination of the study, the monitoring of visits to each centre, data management, data entry, and data analysis were conducted by the Grenoble clinical investigation centre (university of Grenoble, France), which is independent from the evaluated spa centres.

1.3. Results

1.3.1. Patients

A flow chart of the patient analysis procedure is presented in Fig. 1. Four hundred and sixty-two patients were recruited into the original study but 11 refused both groups after Zelen randomisation leaving 451 enrolled into the original study. From these 301 directly reported the side of the knee affected by OA on the questionnaire and an additional 98 had swelling and/or crepitation and/or effusion in one or two knees, and so 399 patients were analysed for GOA. The remaining 52 patients were considered to have unilateral knee OA and not included for further analysis. Of the 399 patients assessed for inclusion in this GOA trial, 70 (15.52%) had Heberden's nodes, 185 (41.02%) had fulfilled the ACR criteria, and 185 (41.02%) displayed the Dougados criteria. In total, 214 (47.45%) patients fulfilling at least one of the three definitions of GOA and were included in the post-hoc analysis. This sample consisted of 101 patients from the original control group and 113 from the spa therapy group. At baseline the groups were similar for all variables except for body mass index (BMI), WOMAC pain score, WOMAC function score, VAS pain, and SF36 PCS (P < 0.05). Table 1 shows criterion values at baseline.

1.3.2. Outcomes

Minimally clinically important improvement was observed more often (P = 0.010) in the spa group (52/94; 55.3%) than in the control group (32/88; 36.4%). No surgery was performed in these patients. The adjusted odds ratio was 1.99 ([1.06–3.77]; P = 0.033) after the adjustment of the three outcomes with significant differences at baseline (WOMAC pain, WOMAC function, and pain) and the step-down model; the NNT is 5.3. The results for the main outcomes and MCII are presented in Table 2. Secondary outcomes are reported in Table 3.

For each criterion, we tested the interaction between the group and GOA. No significant results were obtained. For the primary criterion, the rate of patients with MCII at six months was 36.4% in the control group vs. 55.3% in the treated group for patients with GOA. For patients without GOA, the rate was 36.4% in the control group vs. 46.5% in the treated group (no significant interaction; P = 0.400).

1.3.3. Harms

Several side effects were reported from the spa therapy group, venous insufficiency (1), acute low back pain (2), erysipelas (1), asthenia (1), and joint effusion (1). No side effects were recorded in the control group.

1.4. Discussion

Our data seems in favour of spa therapy for patients with GOA. This post-hoc subgroup analysis of a randomised study indicate that spa therapy in combination with a home exercise programme may provide better clinically relevant improvement in pain and knee function in comparison with a home exercise programme alone in patients suffering from GOA. However, the improvements in patient acceptable symptom state or quality of life were not sufficient to reach a significant improvement.

We acknowledge that this subgroups analysis does not fulfil all the criteria for subgroup analysis as described by Sun et al. [31]. However, these were developed to prevent the development of false-positive results by chance when the original study results are negative whereas our original study reported a positive effect. In addition, we discovered that most of the patient had no knee OA but GOA after the conclusion of the study, so it was not possible to plan this analysis in the protocol. Further, we believe that the publication of this subgroup analysis is justified by the fact that the localized knee osteoarthritis and generalized osteoarthritis are two different pathologies.