This was a randomized, double-blind, placebo-controlled clinical trial performed in Rheumatology Clinic of Al-Zahra hospital of Isfahan, Iran, affiliated to Isfahan University of Medical Sciences (IUMS), from October 2020 to April 2021. The study was registered in Iranian Registry of Clinical Trials (IRCT) with the record number of IRCT20150721023282N14.
Preparation of oily lotions. Topical formulation of frankincense extract was prepared by Fardis pharmaceutical company, Isfahan, Iran. For preparing 100 ml of Boswellia lotion, 1 g of dried extract of frankincense was added to 20 ml of black seed oil, then reached to the volume with olive oil. The placebo lotion was prepared just with 20 ml of black seed oil and 80 ml of olive oil. Both types of lotion were similarly packaged in pharmaceutical tubes and labeled and a numeric code was recorded on each tube by the company.
Patient selection. Patients were selected from those referring to rheumatology clinic of Al-Zahra hospital, based on the following inclusion criteria: (1) age of 40–80 years, (2) OA of at least one knee for at least 3 months based on the diagnostic criteria of American College of Rheumatology (ACR), (3) pain score > 4 based on Visual Analogue Scale (VAS), and (4) grade 2 or 3 of Kellgren-Lawrence scale in knee radiography within the past 3 month.
The patients’ exclusion criteria were: (1) use of intra-articular glucocorticoids within the past 3 months, (2) use of intra-articular sodium hyaluronate within the past 6 months, (3) use of systemic glucocorticoids (either oral or parenteral) within the past 14 days, (4) concurrent other osteoarticular disorders (e.g., rheumatoid arthritis and gout), (5) any skin disorder in the knee region, (6) any allergic reaction to the prescribed topical preparation, (7) knee arthroscopy within the past 3 months, (8) illiteracy, (9) inability to answer the questions, and (13) pregnancy or lactation (for women).
Clinical study and interventions. All included participants filled a written informed consent form. Before any intervention, the pain severity based on VAS (0 to 10 scale), and the scores of WOMAC (Western Ontario and McMaster Universities Osteoarthritis Index) and PGA (patient global assessment) were determined and recorded for the patients according to the symptoms experienced within the past 48 hours. WOMAC index consists of three subscales including pain (5 questions), the flexibility of knee joint (2 questions), and the function or daily activities (17 questions) that a score range of every question presents the severity of the symptom. The total score of WOMAC ranges from 0 to 68, as the best and worst state for knee osteoarthritis, respectively. The Persian form of WOMAC used in this study, has been validated previously by Nadrian et al . PGA is a visual measurement scale in the form of a colored band that ranges from white zone/point 0 (without any symptoms) to black zone/point 4 (the worst symptoms) and it was marked based on patient self-assessment. The tubes containing drug or placebo were equally and randomly given to eligible patients with the code being written on the data collection forms. Patients were asked to apply the lotion on the involved knee, three times daily for 4 weeks. Also, acetaminophen 500 mg three times a day was prescribed for all patients, as a standard treatment for OA. Furthermore, the patients were asked not to use any other drug/supplement for OA, including glucosamine, chondroitin, herbal medicines, and topical preparations during the study period. The patients were instructed to record regular use of the lotion and report any side effect during the study. As the tubes of drug and placebo lotions were fully similar, the prescribing physician (rheumatologist), the data collector, and the data analyst were all blinded to the type of intervention (drug vs. placebo) for each patient. At the end of the study, VAS, WOMAC, and PGA scores were recorded again for all patients.
The primary outcome measures were the changes of VAS, WOMAC, and PGA scores at the end of study. The secondary outcome variable was the rate of possible side effects based on the report of the patients.
Sample size calculation. The following equation was used for sample size calculation:
n = (Z1−α/2 + Z1−β)2 × 2𝛿2/(µ1 - µ2)2
Where n is the required sample size in each group; µ1 and µ2 are the mean of the variable in the first and second groups, respectively, according to the previous studies; 𝛿 is the standard deviation (SD); Z1−α/2 is the standard normal z-value for a significance level α = 0.05, which is 1.196, and Z1−β is the standard normal z-value for the power of 80%, which is 0.84. According to the µ and 𝛿 values for pain scores (in WOMAC) in a previous report , a calculated sample size of at least 25 patients was considered for each group.
Statistical analysis. Data analysis was performed by SPSS 24.0 software (SPSS Inc., Chicago, IL, USA). Qualitative variables were compared by Chi square (χ2) test and Fishers’ exact test between the two groups. Kolmogorov–Smirnov test was used to determine distribution pattern of quantitative data. The normally and non-normally distributed data were presented as mean (SD) and median [IQR], respectively. Wilcoxon Signed Rank test and Mann- Whithney U test were applied for comparison of values at the beginning and end of intervention within each group and between the groups, respectively. P-value < 0.05 was considered as statistically significant.