This randomised trial of krill oil for osteoarthritis of the knee (KARAOKE) study is a multicentre, randomised, double-blind, placebo-controlled trial over 24 weeks. The trial was registered on the Australian New Zealand Clinical Trials Registry prior to recruitment, and trial reporting will be guided by the Consolidated Standards of Reporting Trials (CONSORT) Statement. We aim to recruit a convenience sample of 260 patients with clinical knee OA, significant knee pain and effusion-synovitis visualised on MRI. Patients will be recruited via the OA Clinical Trial Network, at Australian public hospitals in Melbourne, Sydney, Adelaide and Perth and a research institute in Hobart, using a combined strategy, including collaboration with general practitioners, rheumatologists, and orthopaedic surgeons, as well as advertising through local and social media. Patients will be encouraged to contact their local research nurse via email or telephone.
Inclusion criteria are as follows: males and females aged ≥ 40 years old; with significant knee pain on most days (defined as a pain score ≥ 40 mm on a 100-mm VAS); and meeting the American College of Rheumatology (ACR) criteria for symptomatic knee OA, assessed by a physician, and any effusion-synovitis present on MRI (defined as ≥grade 1 according to modified WORMS scoring[18, 38] ;
Exclusion criteria were as follows:
Significant knee injury within the last 6 months.
Use of anticoagulants, high dose aspirin or non–steroidal anti–inflammatory (NSAIDs), as krill oil is contra–indicated in such people.
Unwillingness to stop taking krill oil and fish oil medications 30 days prior to the trial and during the trial (minimum washout 4 weeks).
Other forms of inflammatory arthritis (especially rheumatoid arthritis and gout).
Arthroscopy or open surgery in the “study” knee in the last 12 months
Injections of corticosteroids (last 3 months) or hyaluronic acid (last 6 months) in the index knee.
Women who are pregnant or breastfeeding
Use of any investigational drug(s) and/or devices within 30 days prior to randomisation
Presence of any serious medical illness that may preclude 24 week follow up.
Inability to provide informed consent.
Inability to have an MRI (claustrophobia, pacemakers, metal in eyes, metal in knees that disrupt the images at the area of interest).
Severe knee OA (joint space narrowing (JSN)) on X-ray of Grade 3 using the Osteoarthritis Research Society International (OARSI) atlas).
Randomisation and blinding
Study participants were allocated to receive either krill oil or placebo in a 1:1 ratio based on computer generated random numbers using a central randomisation website hosted by the University of Tasmania using adaptive allocation (minimisation). Briefly, the randomisation program examined the number of participants that were currently assigned to each of the two arms and then adjusted the randomisation thresholds so that the arm with the fewest participants would have a greater chance of being selected. Randomisation will be stratified by study site and is accessible by a dedicated website. This will be conducted by staff members with no direct involvement in the study.
The randomised controlled trial will be a double-blind one, with both patients and investigators assessing outcomes blinded to treatment allocation. Allocation concealment and double blinding will be ensured by 1) use of identical softgels for each group; 2) objective measures of knee structural changes being made by trained observers blinded to group allocation; and 3) subjective measures being taken by research nurses blinded to group allocation.
Emergency unblinding will be allowed in limited situations that impact on the safety of study patients. Code-break for the full randomisation schedule will be maintained by the University of Tasmania. Patients who are unblinded will be withdrawn from treatment but will continue to be followed as per the planned follow-up schedule.
Eligible people will receive 2 x 1g softgels daily of either krill oil or placebo for six months. The krill oil (Superba Boost product) contains 190 mg/g of EPA and 100 mg/g of DHA. The total omega-3 content is 350 mg/g, total omega-6 is 12 mg/g, hence the omega-3 to omega-6 ratio is 29. The placebo used is a mixture of vegetable oils (virgin cold pressed olive oil, maize oil, palm kernel oil, medium chain triglycerides), containing no EPA or DHA, and less than 5 mg/g (0.5%) other omega-3s (predominantly oleic acid (C18:1n9) and linoleic acid (C18:2n6)). Both the krill oil and the placebo are provided in non-distinguishable opaque glycerin softgels. A small amount of vanilla flavour added to the shell of both active and placebo softgels to ensure uniform taste and smell. All patients will continue usual care by their treating health practitioners.
Study procedure and time points
Research assistants will first conduct screening over the telephone. If early checks of study eligibility are favourable, participants will be booked in for a face-to-face screening visit to further determine eligibility and explain what is involved in the study. At the screening face-to-face visit, patients will complete questionnaires, have a knee x-ray and MRI, supply a blood sample, and have a clinical assessment by a study doctor to ensure inclusion criteria are met. The study knee will be defined as the one with symptomatic OA meeting all inclusion criteria. If both knees meet these criteria, the study doctor will decide which is the study knee, typically the one with the highest pain score.
Table 1 outlines the schedule of assessments. After screening, there will be 3 study visits (week 0, 12, 24). The same research assistants, who are blinded to treatment allocation, will measure all clinical variables, administer questionnaires, monitor compliance, and record adverse events at these visits. Additional questionnaire mail outs will occur monthly. MRI scans will occur at screening and week 24; knee x-ray will be performed at screening; blood samples are taken at screening, 12 and 24 weeks, and urine samples are taken at baseline and week 24.
Participants will discontinue involvement in the study if they need to use drugs that are contraindicated (marine oils, anticoagulants, high dose aspirin, NSAIDs), they stop taking study medication, or the participant or the site investigators request discontinuation.
To ensure high-quality execution of the trial in accordance with the protocol, all trial staff will be trained by the chief investigators and provided with a standard protocol book which contains details of standard operating procedures, trial contacts, visits, measurements, monitoring and case report forms. Data is collected using the same forms across sites, this data is checked by staff in Hobart and any discrepancies are clarified.
We have two primary outcomes for this study, assessing improvements in symptoms and structural progression. These are change in knee pain, as assessed by VAS over 24 weeks and change in effusion-synovitis volume assessed from MRI. All outcomes and time points of assessment are listed in Table 2.
Other Outcome measures
Knee pain, back pain and hand pain will be assessed using a 100mm VAS by asking “on this line, thinking about your study knee / back / most painful hand, where would you rate your pain, using the last seven days as a time frame” over 4, 8, 12, 16 and 20 (but not 24) weeks.
We will also assess pain using the Western Ontario and McMasters Universities Osteoarthritis Index (WOMAC), also over the preceding seven days. Five items of WOMAC pain scale in 100-mm VAS format will be used to assess pain during walking on a flat surface, going up and down stairs, at night while in bed, sitting or lying, and standing upright during the last seven days. Items will be summed to create a total WOMAC pain score (range, 0-500). Incomplete items will be addressed according to the WOMAC user guide. If there is one missing item only, the remaining four items will be averaged and then multiplied by five. The WOMAC pain score will be considered invalid if there is more than one missing item.
Knee function will be assessed using WOMAC. Seventeen items of the WOMAC function scale in 100-mm VAS format will be used to assess function during descending stairs, ascending stairs, rising from sitting, standing, bending to floor/picking up an object, walking on flat surface, getting in/out of the car, going shopping, putting on socks/stockings, rising from bed, taking off socks/stockings, lying in bed, getting in/out of the bath, sitting, getting on/off the toilet, heavy domestic duties, and light domestic duties during the last seven days. Items will be summed to create a total WOMAC function score (range, 0-1700). If there are ≤2 missing items, the remaining items will be averaged and then multiplied by 17.The WOMAC function score will be considered invalid if there are >2 missing items.
Magnetic resonance imaging outcomes
An MRI scan of the “study” knee will be performed (screening, week 24). Knees will be imaged in the sagittal plane on a 1.5T or 3T whole-body magnetic resonance unit using a dedicated knee coil. Sequences will include T2-weighted fat saturation 3-dimensional (3-D) fast spin echo sequence (effusion-synovitis volume, cartilage defects, BMLs); T1-weighted fat saturation three-dimensional (3-D) gradient-recalled acquisition sequence (cartilage volume, cartilage thickness); and T2 mapping (see Table 3).
Effusion-synovitis is defined as the presence of intra–articular fluid–equivalent signal on T2 weighted MRI sequence (see Figure 1). A modified Whole-Organ Magnetic Resonance Imaging Score (WORMS) scoring system will be utilised to assess effusion-synovitis (grade 0 to 3) semi quantitatively in 4 regions of interest (ROI; suprapatellar pouch, central portion, posterior femoral recess, subpopliteal recess) in terms of the estimated maximal distention of the synovial cavity. The greatest score for any of the four ROI will be used as maximal effusion of the knee. The intraclass reliability assessed as weighted κ in 50 randomly selected images was 0.63–0.75 in different subregions, and the interclass inter-rater reliability was 0.65–0.79.
Effusion-synovitis volume will be measured using a semiautomated segmentation method according to the intra-articular fluid-equivalent signal on a section-by-section basis in each ROI. The final 3-D volume rendering will be generated using commercial in-house imaging software. Change in effusion–synovitis volume is calculated by subtracting baseline values from follow–up values. The intra-rater reliability was 0.97 in the whole joint (0.98 in suprapatellar pouch and 0.95 in central portion). The inter-rater reliability was 0.99 in the whole joint (0.99 in suprapatellar pouch and 0.93 in central portion). Effusion-synovitis volume was highly correlated with effusion-synovitis score (rho 1⁄4 0.77, P < 0.01 for total, rho 1⁄4 0.91, P < 0.01 for suprapatellar pouch, and rho 1⁄4 0.77, P < 0.01 for central portion).
Differentiating effusions from synovitis on MR images requires administration of a contrast agent (gadolinium), which may cause nephrotoxicity. While this is rare, we do not consider the benefits of such differentiation between synovitis and effusion justifies the risk to our study participants, and hence we have not used it.
Bone marrow lesions
BMLs will be defined as an ill-defined hyperintensity area in the subchondral bone, on MRI. BMLs will be assessed on the sagittal T2 weighted sequences at the medial tibial, medial femoral, lateral tibial, lateral femoral and patella sites using modified WORMS scoring system. The maximum size of each lesion will also be measured using software cursors applied to the greatest area of the lesion, as previously described. We have demonstrated an intraclass correlation coefficient (ICC) of 0.97, using this method. Total BML size will be calculated as the sum of every lesion within the medial tibial, medial femoral, lateral tibial, lateral femoral and patella sites.
Knees will be imaged using an Esaote Mylab 70 VXG ultrasound machine in Perth (see Table 4), with scans undertaken by a single sonographer. Images will be stored to be re read to determine intra reader reliability).
Synovitis in the supra patella pouch will be defined as a combination of synovial hypertrophy and effusion ( using Outcome Measures in Rheumatology (OMERACT) definitions) and scored on a 0-3 scale with grade 0 = no synovitis, grade 1= minimal distension of the recess by abnormal internal hypoechoic or anechoic (relative to subdermal fat tissue) material, grade 2= moderate distension or enlargement of the recess by abnormal internal hypoechoic or anechoic (relative to subdermal fat tissue) material with flat or concave superficial limit, and grade 3= severe distension or enlargement of the recess by abnormal internal hypoechoic or anechoic (relative to subdermal fat tissue) material with bulging superficial limit. Depth of effusion (mm) will be measured only in the suprapatellar pouch (SPP), from a still midline longitudinal image, measuring the maximal effusion depth in this plane.
In the SPP, medial parapatellar region and lateral parapatellar region, synovial hypertrophy, effusion, and power doppler signal (defined using OMERACT definitions) will be scored as absent (0) or present (1).
Pressure pain threshold testing (PPT)
Pressure pain testing (PPT) will be conducted using a pressure algometer (FDN200, Wagner Instruments, USA) with a probe area of 1 cm2, that exerts pressures of up to 200N/cm2/2000kPa. PPT is determined using an ascending stimulus intensity (slowly increasing ramps of 50kPa/s (~0.5kg/cm2s) applied bilaterally over the three test sites: the affected joint (knee, assessed over the tibial below the medial joint line), an unaffected joint (ipsilateral proximal radioulnar joint), and over the thenar eminence. The study participant is asked to notify the examiner as soon as the sensation becomes painful, and the pressure exerted recorded. The PPT is determined by the mean of the threshold from a series of three stimulus intensities each applied as a slowly increasing ramp of 50 kPa/s .
OMERACT–OARSI responder criteria
Response to krill oil will be assessed using the OMERACT – Osteoarthritis Society International (OARSI) set of response criteria. Participants will be classed as responding if they have high improvement in pain (using the VAS) or function (using WOMAC function scale) of ≥50% and absolute change ≥20; or if they have improvements in relative (≥20%) or absolute (≥10) change in 2 out of 3 of pain, function, or patient’s global assessment.
Lower limb muscle strength
Lower limb muscle strength is a key correlate of pain and typically increases when pain reduces. We will assess lower limb leg strength in both legs simultaneously, using a dynamometer (TTM Muscular Meter, Tokyo, Japan) as previously described. This is done in duplicate and the mean of the two estimates is used. The muscles measured in this technique are mainly the quadriceps and hip flexors. The previously published repeatability estimate (Cronbach’s α) for this method is 0.91.
Measure of intermittent and constant osteoarthritis pain (ICOAP)
ICOAP is a multidimensional OA-specific measure designed to comprehensively evaluate the pain experience in people with hip or knee OA. This includes pain intensity, frequency, and impact on mood, sleep, and quality of life, independent of the effect of pain on physical function, over the past 7 days. It contains eleven items in two domains with two supplementary items on intermittent pain predictability, and are rating scales with 5 categories of response eg “not at all,” “mildly,” “moderately,” “severely,” and “extremely”. Each ICOAP item is scored from 0–4; missing data is dealt with according to rules in the User Guide. Scores are produced for each subscale separately by summing subscale scores for each item and then normalising each score from 0 (no pain) to 100 (extreme pain). The ICOAP is reliable (Cronbach’s α 0.93, ICC 0.85), and valid.
Quality of life
Quality of life will be assessed using the Assessment of Quality of Life (AQoL-6D) questionnaire, which assesses six separately scored dimensions (Independent Living, Relationships, Mental Health, Coping, Pain and Senses), each with variable item numbers and response levels. The AQoL-6D has good psychometric properties. Utility scores will be calculated based on methods published on the AQoL website with a range 0-1 where 0 indicates the worst health state and 1 the best.
Fasting blood samples will be obtained from study participants at screening, 12 and 24 weeks. Samples will be analysed for fasting glucose, lipids (total, HDL and LDL), and hsCRP. The blood will be either tested fresh or stored at -80°C.
A standing anteroposterior semiflexed radiograph of the ‘study’ knee will be performed at screening. X-rays will be scored for joint space narrowing and osteophytes on a four point scale (0–3) using the OARSI atlas. In our hands this method has very high reproducibility with an ICC of 0.98 for joint space narrowing and 0.99 for osteophytes.
These include height (stadiometer) and weight (electronic scales) and body mass index (BMI) (weight/height2) measured at weeks 0, 12 and 24.
Neuropathic pain will be assessed by the painDETECT questionnaire (-1 to 38) at screening to provide information on level of nociceptive and / or neuropathic contributions to pain. A painDETECT score <12 is defined as unlikely neuropathic pain, and 13-18 as possible neuropathic pain.
Use of pain medicines will be recorded by questionnaire at all visits. Participants will be asked to keep medications as stable as possible (including NSAIDs) and use paracetamol as rescue medication. Use of fish and krill oil during the trial will not be permitted, and use must cease 2 weeks prior to randomisation. Participants who commence anti–coagulant therapy will be withdrawn from the trial. We will assess analgesic use from the medication data.
Treatment guessing and adherence
Study participants will be asked what treatment they think they received at the 12 and 24 week assessments with the following options: krill oil, placebo, or not sure.
Adherence to treatment will also be assessed at the 12 and 24 week assessments by standard pill count methods.
Adverse events will be monitored throughout the study. Standard safety and efficacy monitoring will be performed through regular face-to-face visits and phone calls between visits. The patients are requested to report any adverse events to the research staff spontaneously. Details of the adverse event and its relationship with study intervention will be recorded and reported to the local Human Research Ethics Committees in accordance with the requirements of individual committees.
Sample size calculations
Using data from another trial conducted in our centre, and assuming 10mm difference between krill oil and placebo on the VAS pain scale (reduction in VAS pain scores in the placebo group by -15.5±25.5 mm over 12 weeks), and using assumptions of 90% power, and 5% probability of type 1 error (alpha=0.05), we will need 234 participants. Adjusting for 10% loss to follow up, we need 260 participants (130 in each arm).
Based on data on effusion-synovitis volume in our vitamin D RCT for knee OA, there was 13.7ml (SD=10.7) in the placebo group and 13.6ml (8.1) in the active group (in those with effusion-synovitis at enrollment; i.e., 60%) at baseline. With 234 subjects we will have 90% power to detect a difference of 4.5ml. It is likely we will see at least a reduction of this amount with a treatment that effectively targets inflammation. What level reflects clinical significance is uncertain but this level is approximately 2.5 times the change that could be expected with measurement error. Furthermore, modelled data from our long-term Tasmanian cohort demonstrates that a reduction in effusion-synovitis size by 4.5ml will decrease the need for joint replacement by 30% over 13 years (unpublished data).
The primary analyses will be intention-to-treat analyses of primary and secondary outcomes. Per protocol analyses will be performed as the secondary analyses.
Changes in knee pain, knee effusion-synovitis size and any other outcomes which were collected using a linear scale will be analysed using a linear mixed model with treatment, month and their interaction (treatment ´ month) as covariates, as well as outcome measured at baseline and the baseline interaction with month. Incidence of adverse events will be assessed using log binomial regression. Correlated data within trial centres and the repeated measures will be addressed using trial centre and patient identification as random intercepts. Month will be treated as random effect to allow different treatment effects among patients over time. The sensitivity of models to the structure for the random effects and covariance structures will be assessed using likelihood ratio tests. Change in outcome measures within each group and differences in changes between groups from baseline to follow-up will be calculated using linear combinations of the estimated coefficients. If there are baseline imbalances in covariates between treatment groups, we will consider adjusting for them based on whether we regard the imbalance as clinically significant. Missing data caused by loss to follow-up and nonresponses will be addressed by adding variables that have complete data at baseline and can explain missingness to the regression models.
Secondary analysis for missing data will be performed in people with full medication adherence (eg >80% of softgels), and using imputation. Baseline variables with complete data will be used for data imputation assuming missing at random.
Subgroup analyses will be performed to examine which subgroups may respond better to treatment. These will include effusion size, radiographic knee OA, co-pathology present on MRI, pain characteristics, levels of inflammatory markers, serum lipid measures. Statistical significance will be set as a two-sided P value <0.05.
Data integrity and management
Data will be recorded using case report forms and processed centrally at the Menzies Institute for Medical Research, University of Tasmania. The hard copies of case report forms will be stored in a locked area at each study site with secured and restricted access. The electronic data will be stored on password protected servers with restricted access. All data collected will be kept strictly confidential. Daily back ups of all electronic data will occur to minimise any risk of lost data. Data transfer will be encrypted with all data de-identified. Only members of the research team who need to contact study patients, enter data or perform data quality control will have access to patient information.
After study completion, paper copies of data will be archived in secure storage. Identifiers will not be removed, in case follow-up of study patients is necessary; however, electronic data will continue to be kept in a secure electronic database. This will remain password protected and with access given only to the study investigators unless otherwise authorised by the study team.
If patients withdraw from the study before 6 months of follow-up, the reason and date will be recorded. An early MRI will be considered for participants who can not make the final visit.
Roles and responsibilities and monitoring
The University of Tasmania (as the trial sponsor) and the principal investigators are responsible for all aspects of the trial, including design, conduct and oversight. The principal investigators will monitor the conduct and progress of the project at each site. The trial coordinator will visit each study site to make sure that all trial procedures are compliant with the trial protocol. The principal investigators and the research team will have regular teleconferences to ensure efficient study execution and ongoing monitoring of the study progress, with summary documents circulated after each meeting. Krill oil has a good safety profile, and is available over–the–counter, so we do not plan to use a data safety monitoring board. The trial is also being monitored at each site by a practicing rheumatologist.
The results of this study will be presented at conferences and published in scientific journals. Any notes or publications arising from our research will be de-identified. Only aggregate statistical results will be presented.
The outcomes of the project will be disseminated to study patients using non-technical language. Dissemination of the overall study findings to the patients will occur in a de-identified manner and be based on the entire study population. The scientific paper will be available for dissemination to study participants.