Clinical and technical evidence

A literature search was carried out for this briefing in accordance with the published process and methods statement. This briefing includes the most relevant/best available published evidence relating to the clinical effectiveness of the technology. The literature search strategy, evidence selection methods and detailed data extraction tables are available on request by contacting mibs@nice.org.uk.

Published evidence

Evidence in this briefing is taken from 1 meta-analysis and 4 randomised controlled trials (RCTs). The total number of patients included in the 4 RCTs was 3,213. The results varied among the studies.

Table 2 summarises the clinical evidence as well as its strengths and weaknesses.

Table 2 Summary of selected studies

Study  

Details of intervention and comparator

Outcomes  

Strengths and limitations

Conaghan et al. 2013

1,395 patients, randomised, double-blinded, placebo and active-controlled trial.

Multicentre (Czech Republic, Germany, Poland, UK).

IDEA-033 (ketoprofen in ultradeformable vesicle gel; not a marketed product).


Oral celecoxib.


FLEXISEQ (TDT 064) (ketoprofen-free topical placebo).


Oral placebo.

When using the WOMAC index to assess pain, IDEA-033 was not superior to FLEXISEQ in reducing osteoarthritis knee pain.

IDEA-033 and FLEXISEQ were superior to oral placebo and non-inferior to celecoxib in reducing osteoarthritis knee pain.

The most frequent types of treatment-related adverse events reported were gastrointestinal symptoms for oral (celecoxib, oral placebo) and skin reactions for topical applications (IDEA-033, FLEXISEQ).

The study was a randomised controlled trial (RCT), which is a strength of the study design.

FLEXISEQ was designed to be the topical placebo gel arm so there was no comparison of FLEXISEQ with a true topical placebo (such as an inactive gel) which would have controlled for any placebo effects of administering a topical treatment, such as patient expectation of benefit. The study reported a detailed sample size calculation and justification.

As with all patient-reported primary outcomes, the WOMAC score may be subject to bias (this scale is commonly used to assess the condition of patients, based on their symptoms and physical functioning, it measures 5 items for pain [score range 0 to 20], 2 for stiffness [score range 0 to 8], and 17 for functional limitation [score range 0 to 68]). The WOMAC index is the recommended methodology for assessing efficacy in osteoarthritis trials (Dieppe, 1995).

The use of the baseline-observation-carried-forward (BOCF) approach to handle missing data could have resulted in an underestimation of the treatment effects.

The study received funding from the manufacturer of IDEA-033 which may introduce bias.

Kneer et al. 2013

866 patients, randomised double-blind placebo-controlled parallel group,

multicentre (Germany, Poland, Serbia, Croatia).

IDEA-033.


FLEXISEQ (topical placebo).

All 4 treatments improved WOMAC pain scores by 50% or more.

The 50 and 100 mg IDEA-033 doses (but not the 25 mg) were marginally superior to FLEXISEQ in reducing pain. The superiority was not demonstrated for the WOMAC function score.

The percentage of patients who responded was significantly higher for all the IDEA-033 groups versus the FLEXISEQ group.

Skin reactions were the only relevant drug-related adverse events. No significant differences were reported between the groups in terms of adverse event frequency.

RCT with blinding. However, as FLEXISEQ was designed to be the topical placebo gel arm, there was no comparison of a true topical placebo with FLEXISEQ.

The study reported a detailed sample size calculation and justification.

As a patient-reported outcome, the WOMAC score may have been subject to bias.

A large number of patients in the IDEA-033 groups withdrew from the study at an early stage and this may have contributed to the lower than expected effect size in these treatment groups. Equally the larger number of drop-outs could reflect dissatisfaction with either effectiveness or side effects.

The study received funding from the manufacturer of the IDEA-033.

Rother and Conaghan 2013
555 patients, randomised double-blind trial,
multicentre (USA).

IDEA-033.


FLEXISEQ (topical placebo).

IDEA-033 was inferior to FLEXISEQ in relieving mild-to-moderate pain associated with knee osteoarthritis and improving joint function.

IDEA-033 was associated with a higher frequency of withdrawals due to adverse events compared to FLEXISEQ.

RCT with blinding. However, as FLEXISEQ was designed to be the topical placebo gel arm, there is no comparison of a true topical placebo and FLEXISEQ.

The study reported a detailed sample size calculation and justification.

As a patient-reported outcome, the WOMAC score may have been subject to bias. The study received funding from the manufacturer of IDEA-033 which may introduce bias.

Rother et al. 2007

397 patients, randomised, double-blind, controlled trial,

multicentre (Germany).

IDEA-033.


Oral celecoxib.


FLEXISEQ (topical placebo).


Oral placebo.

IDEA-033 was superior to placebo and comparable to celecoxib in relieving pain associated with an acute flare of knee osteoarthritis.

Most adverse events were evenly spread throughout the groups, but IDEA-33 caused more skin irritations than FLEXISEQ.

RCT with blinding. However, as FLEXISEQ was designed to be the topical placebo gel arm, there is no comparison of a true topical placebo and FLEXISEQ.

The study reported a detailed sample size calculation and justification.

As a patient-reported outcome, the WOMAC score may have been subject to bias.

The study received funding from the manufacturer of IDEA-033 which may introduce bias.

Rother et al. 2014

Meta-analysis of 5 RCTs with FLEXISEQ using formal parametric meta-analysis methods to investigate the overall effect size (ES) of FLEXISEQ across all studies, versus the standard placebo effect (pooled data of 198 osteoarthritis trials).

Intervention: FLEXISEQ (5 RCTs, n=1,320).


Comparator data from Zhang et al. (2008): all placebo-controlled knee osteoarthritis groups (122 trials, n=10,300); all placebo groups for topical NSAIDs (13 trials, n=896); all untreated controls (14 trials, n=1,167).

The combined ES for FLEXISEQ for pain relief and improvement in function was higher than the values reported for placebos in knee osteoarthritis studies, and for the placebo arms reported in topical and oral NSAIDs studies.

The effect sizes for the individual FLEXISEQ studies were higher than the effect size reported by Zhang et al. (2008).

Funding for the 5 RCTs in this study was provided by the manufacturer of IDEA-033; funding for the meta-analysis was provided by the manufacturer of FLEXISEQ.

Four of the 5 RCTs in the meta-analysis have been published as peer reviewed papers. The fifth RCT was reported as a conference abstract.

The meta-analysis seemed to be of poor quality. The search strategy was not reported and it was unclear what process was adopted for sifting and data extraction. For the included studies, no assessment of study bias is reported, with the exception of conflicts of interest.

Abbreviations: baseline-observation-carried-forward (BOCF) approach; RCT, randomised controlled trial; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index.

Strengths and limitations of the evidence

All 4 identified studies were RCTs and included a detailed sample size calculation and justification. The majority of the data on the efficacy of FLEXISEQ are derived from the clinical development programme for IDEA-033 which included RCTs with FLEXISEQ as the drug-free control (IDEA-033 is a topical gel in development, which contains ketoprofen in Transfersome vesicles, using the same core technology as Sequessome vesicles; Pro Bono Bio, Transferosomes, 2016). This means that there was no placebo topical treatment with an inactive gel that would control for any placebo effects of administering a topical treatment, such as patient expectation of benefit, making this a major limitation in all 4 RCTs.

The risk of selection bias in Kneer et al. (2013) was unclear. The paper did not report the method of randomisation, and it was unclear whether there was adequate concealment of allocation for investigators and patients. The risk of selection bias in the other 3 studies was deemed to be low, as either a computer-generated list or a random permuted block scheme was used to randomise patients into groups. In all 4 studies, the treatment groups were comparable at baseline, with only Kneer et al. (2013) reporting that they did a test on statistical significance.

None of the studies was deemed to be at high risk of performance bias because in each study, the treatment groups had the same care other than the intervention. All 4 studies are reported as double-blind, but it is not clear how patients were kept blind to their treatment.

All 4 RCTs had 12-week follow-up, which is typical for studies of treatments for osteoarthritis. A Cochrane review of topical NSAIDs for chronic musculoskeletal pain (April 2016) identified 39 studies that were generally of high quality, none of which lasted longer than 12 weeks.

In Kneer et al. (2013), the effects of FLEXISEQ administered alone were evaluated through the 'NSAID-responder enrichment design', which involves enrolling patients experiencing an increase in symptoms (flare) following withdrawal of NSAID treatment. This study design has been used in studies of COX-2 inhibitors, and has been associated with a greater treatment effect than non-enrichment designs (Trijau et al. 2010; Bjordal et al. 2007). Therefore, the selection of patients likely to respond to NSAIDs in this study may have contributed to the statistical superiority of the IDEA-033 groups compared with FLEXISEQ. In the other 2 phase III trials (Conaghan et al. 2013 and Rother and Conaghan 2013), a non-flare design was used and these showed stronger FLEXISEQ effects.

Only 1 of the studies included a UK centre (Conaghan et al. 2013) which may therefore limit the generalisability to the NHS.

All 4 studies received funding from the manufacturer of IDEA-033 and the manufacturer of FLEXISEQ provided publication support for the meta-analysis. Manufacturer-sponsored studies may introduce bias into publication results and conclusions.

Rother et al. (2014) used a network meta-analysis to compare the effect size of FLEXISEQ (from 5 RCTs) with the placebo effect, using a large dataset of pooled placebo response results from a systematic review and meta-analysis of 198 RCTs on osteoarthritis (including 193 placebo groups; Zhang et al. 2008). The meta-analysis seemed to be of poor quality. Four of the RCTs that involved FLEXISEQ were published in peer reviewed journals and 1 was reported as a conference abstract. Although the methodology used by Rother et al. (2014) is similar to that used in the meta-analysis of RCTs looking at osteoarthritis treatments in Zhang et al. (2008), there are issues with the comparator (pooled placebo effect from the Zhang et al. study) that could potentially introduce bias. Firstly, Zhang et al. (2008) used study level variables for the regression analysis so the sensitivity analysis may have been lower than expected. Secondly, the limited information provided in some of the Zhang et al. studies made it difficult to properly assess their quality before including them in the meta-analysis and pooling the data.

Recent and ongoing studies

One ongoing trial has been identified: