Inclusion criteria.

Eligible patients were male and female subjects aged between 60 and 85 years, with OA of the knee as defined by criteria of the American College of Rheumatology [21], with pain in the affected knee and a confirmatory knee X-ray diagnosis (Kellgren Lawrence grades II-III) [6].

Exclusion criteria.

Exclusion criteria were patients aged less than 60 years or more than 85 years, Kellgren Lawrence grades I and IV, mental or neurologic deficit, recent knee trauma or suspicion of another joint affection, uncontrolled systemic diseases, thrombocytopenia, bleeding tendencies, use of anticoagulants or antiaggregants, and recent myocardial infarction or stroke.

Groups.

Patients were randomized into two groups: ozone group (OZ) and placebo group (PBO). Patients from the OZ group received one intra-articular injection of ozone 20 μg/ml—10 ml [10]. Due to the short half-life of ozone (approximately 45 minutes at 20°C, it was freshly generated in the clinics, using an Ozone & Life O&L 3.0 RM generator (Sao Jose dos Campos—Brazil) connected to a pure oxygen source and used immediately for the patient. Ozone generators use oxygen through high voltage tubes with outputs ranging from 4,000–14,000 and produce an ozone-oxygen mixture with concentration ranges extending to 5% [17]. Placebo group patients received an intra-articular injection of 10 ml of air. Each patient received one injection (OZ or PBO) once a week for 8 consecutive weeks. All patients were treated using the sterile injection technique.

Sample size calculation.

The sample size calculation was determined to guarantee the statistical power mainly for the two primary endpoints. For these variables, a sample size of 40 evaluable patients provided an 80% power to detect a difference of efficacy of 30% between the groups, with a two-sided alpha level of 0.025 and β = 0,20 based on chi squared test [22]. Therefore, a total of 80 evaluable patients were required to analyze the primary endpoints of the study and approximately 96 patients were predefined to be randomized considering a dropout rate of about 20%.

The assortment was not balanced. That occurred because it was made without a control of the total number of patients in each group. Thus, at the end of the study, the relation between the groups was ozone group 1.75 patient: placebo group 1 patient [23]. Because the clinical trial consisted in comparing a new treatment to a pattern, in such a way of acquiring experience and knowledge about the general profile of this treatment, such influences make it necessary to consider the allocation of more than half the sample in this new treatment, even if it occurs some loss of the statistical efficiency [24].

Allocation concealment.

Opaque envelopes containing the group to which each participant would belong were sequentially numbered and closed by an individual not involved in the study. The envelopes were opened by a nurse in a sequential manner after each patient’s evaluation.

Allocation masking.

The syringes containing the treatments were delivered by the nurse to the main researcher in closed packages marked with the patient’s initials.

Patient inclusion in the study.

Study participants attended a baseline visit at which the following procedures were performed: medical history, physical examination, analysis of X-ray of the affected knee and application of the following questionnaires and tests: Visual Analogue Scale (VAS) [25], Lequesne Index [26], Timed Up and Go Test (TUG Test) [27], Short-Form Health Survey (SF-36) [28], Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) [29], and Geriatric Pain Measure (GPM) [30]. Eligible patients were fully informed of the purpose of the study. All patients that fulfilled the inclusion criteria signed the informed consent prior to enrollment in the trial. They were instructed to continue their medical treatment according to their physicians’ orientations.

Randomization.

With the objective of avoiding selection bias, all included participants were sequentially assigned by the researchers to receive OZ or PBO according to a pre-established computer-generated global randomization list. That list was generated by Dr. Fânia Cristina dos Santos (FCS), on November 20th 2010, using software ETCETERA, version 2.46, and constituted 98 numbers with the corresponding treatments (Fig 1). Prior to the beginning of the randomization it was stipulated that group A would be the ozone group and B, the placebo group (Fig 1). Assessments were performed at baseline (visit 1), 4 weeks (visit 2), 8 weeks (visit 3), and 8 weeks after the end of the injections (visit 4). At the follow-up visits, the same procedures as those described for visit 1 were performed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart of the distribution of study patients.

https://doi.org/10.1371/journal.pone.0179185.g001

Researcher masking.

Patients, the main researcher, and researchers that evaluated the outcomes did not know which group the patients were allocated to.

Treatment.

Everyone involved in the study was instructed by the main researcher about the way to generate ozone, as well as the cautions to be taken during the process. A member of the study (among the nurses Iara Monteiro and Maria Sonia Sousa Castro Sant’Ana from Paulista School of Medicine—Federal University of Sao Paulo Geriatrics and Gerontology Discipline Clinic and Luciana Maria Oliveira Bueno de Jesus from Pro-Vida—Center for Total Health Assistance LLC and from Santo Amaro University—Medical College) generated the ozone or placebo—according to the criteria of the Randomization Table and in the order of the closed envelopes—and gave the treatment to the main researcher. The nurses were the only members of the study who knew which treatment each patient received. Neither the patient nor the researcher physician had knowledge of whether the syringes dispensed to each patient contained ozone or placebo as they were identical. The substance that was used as placebo was air because it has characteristics identical to those of ozone (except for the smell). Each nurse carefully placed a needle on the beak of the syringe (containing ozone or air) and put a blood collection bottle cover over the needle. In this way, detection of the ozone smell was avoided, as well as its leakage and consequent loss. The syringes containing ozone were maintained in a vertical position, with their beaks upwards, before they received the needle and blood collection bottle cover. Thus, the leakage of ozone from the syringe was avoided since it has a molecular weight higher than oxygen. Next, the syringes were delivered to the main researcher to be used in the treatment.

Ozone.

The ozone for medical use was obtained from an ozone generator (Ozone & Life—model O&L 3.0 RM, of Brazilian fabrication), composed of a high voltage tube through which medical oxygen (O₂) passes, dividing into molecules that generate ozone.A 10ml syringe was connected to the exit of the generator and 10ml of the produced gas were collected using the following parameters:

In this way an ozone concentration of 20μg/ml was obtained [10]. During the ozone generation process, the room was kept ventilated to facilitate the process of dispersion of the gas that could escape to the environment.

Technique.

The knee that was punctured was the more painful and less functional one. Each patient was positioned sitting over a stretcher with the legs bent. The knee to be punctured was submitted to an antisepsis procedure with gauzes profusely soaked in 70° GL alcohol, using round centrifugal movements from the center of the region to be punctured in the direction to its periphery, 3 to 4 times consecutively. The point of entrance of the needle was the femorotibial articular interline, 1.5cm medially to the patellar tendon and 1.5cm bellow the apex of the patella. In cases when the puncture was not possible using this point of entrance, the femorotibial articular interline 1.5cm laterally to the patellar tendon and 1.5cm below the apex of patella was used. The direction in both cases was strictly anteroposterior to avoid the puncture of the Hoffa’s fat pad [31]. In cases when the needle made contact with with the femoral condyle or when there was an error in the knee puncture, another puncture was made. An anesthetic effect was obtained by injecting 0.5ml of 2% lidocaine solution (without vasoconstrictor), with a 1ml syringe and a 30 x 7mm needle, in the route of the puncture [32]. Care was taken to aspirate the syringe before injecting its content to avoid joint effusion that might be present and ensure that the needle was not inside a blood vessel. This was carried out until it was confirmed that the needle was inside the knee joint. Afterwards, the syringe containing ozone or air was connected to the needle used to anaesthetize the route of the puncture and its contents were administered in a slow and continuous manner. The needle and syringe were withdrawn from the knee joint, the region of the puncture was plugged and a dressing was applied with gauze and micropore tape. The patient was instructed not to remove the dressing for at least 30 minutes after the procedure and avoid making efforts with the punctured joint for at least 24 hours. The procedure was carried out once a week, for 8 consecutive weeks.

Evaluation tools.

Visual Analogue Scale (VAS): VAS is one of the most commonly used instruments to measure pain in the general population as it is considered the most sensitive, reproducible, and simplest pain scale. It is a 10-centimeter line with anchors at both extremities, with the words “without pain” at one end and “unbearable pain” at the other end. The patient is required to mark a point indicating their pain and a 0-100mm ruler is used to quantify the measure [25].

Lequesne Index: Lequesne index comprises 10 specific questions for patients with knee osteoarthritis, 5 related to pain or discomfort, 1 to maximum distance walked, and 4 to daily life activities. The score varies between 0 and 24 points, and the higher the score, the worse the pain and function [26].

Timed Up and Go (TUG test): In TUG test, the patient is required to stand up from a chair (height of seat = 45cm and of arms = 65cm), walk 3 meters, return and sit down again, while the time spent performing the test is timed. The proposition of the test is to evaluate balance when sitting, transferring from a sitting position to a standing position, stability when walking and turning when walking without using compensatory strategies. Independent individuals, without balance alterations, perform the test in 10 seconds or less; with independence for basic transfers, they take 20 seconds or less. Individuals who need 30 seconds or more to finish the test are dependent in many daily life activities and moving, presenting a greater risk of falling [27].

SF-36 Health Survey Instrument: SF-36 is an instrument of generic evaluation of quality of life, characterized by being easy to apply and understand. The questionnaire contains 36 items, divided into eight aspects: functional capacity, physical aspects, pain, general state of health, vitality, social aspects, and mental health. The score for each item of the questionnaire varies between 0 and 100, where zero is equal to the worst state of health and 100 to the best [28].

WOMAC (Western Ontario and McMaster Universities) Index: WOMAC Index contains 24 questions that evaluate pain, stiffness, and physical function during daily life activities (for example, climbing down stairs). The individual is required to indicate the degree of difficulty from 0 (none) to 5 (very strong), of pain and stiffness during the previous 72 hours. The sum of the points given to the 24 items generates a value which varies between 0 and 96; the higher the value, the worse the symptoms of the patient [29].

Geriatric Pain Measure (GPM): GPM was developed to be a multifunctional pain scale, of easy applicability and comprehension to be used in aged populations. It evaluates pain and its impact on mood, daily activities and, predominantly, quality of life. Thus it allows evaluation of the impact of pain on functionality and quality of life in older individuals [30].

Statistical analysis.

The efficacy analysis was performed in the per-protocol (PP) population defined as all randomized patients who met the inclusion / exclusion criteria, received the treatment and from which data from the baseline, 4th week, 8th week, and 16th week visits were available and who did not present major protocol deviations [33]. Major protocol deviations included lack of fulfillment of the selection criteria, voluntary study exit, and non-compliance with the study treatment. The control population included all randomized subjects who received an air injection. Intention-to-treat analysis (ITT) was not carried out for the 2 patients who dropped out the study because they performed only the baseline evaluation.

Data analysis.

Data were organized in an Excel spreadsheet for calculation of score variables (Lequesne Index, WOMAC, and Geriatric Pain Measure) and were analyzed using the statistical program SPSS 20.0. Ages were described according to the groups by using the summary measures median, standard deviation and mean and compared between the groups using the Student’s t test. Qualitative characteristics were described according to the groups using absolute and relative frequencies and the existence of an association between the groups was verified using the Chi-squared test, Fisher’s exact test, or likelihood ratio test, and schooling was compared with Mann-Whitney’s test [34]. Percentage alterations of each scale according to the basal values were created. Scales were described according to the groups and evaluation moments using summary-measures, and comparisons between the groups and moments were performed using generalized estimating equations with autoregressive correlation matrices of order 1 between the moments, with normal marginal distribution and identity or logarithmic link function [35]. For models that presented statistical significance, the analysis was followed by Bonferroni’s multiple comparison test to establish between which groups and scales the differences in the scales occurred [36]. Results were illustrated using medium profile graphics, with the respective standard errors and according to the groups, and the tests used a significance level of 5%.