A 6-months, randomised, placebo-controlled evaluation of efficacy and tolerability of a low-dose 7-day buprenorphine transdermal patch in osteoarthritis patients naïve to potent opioids

2.1 Study medication

Buprenorphine is a semi-synthetic opioid, a derivative of the opium alkaloid thebaine. Buprenorphine has been in regular clinical use for three decades, administered as sublingual tablets or as injections for acute or chronic pain. Pharmacodynamic and pharmacokinetic properties of buprenorphine have recently been thoroughly updated and documented in a review by Kress [8]. The longer half-life of buprenorphine and its high affinity to the μ-receptor, make it longer lasting and 25 to 100 times more efficacious (per mg) as an analgesic than morphine [8]. Buprenorphine is an antagonist at the κ-opioid and the δ-opioid receptors and therefore lacks the psychotomimetic effects and dysphoria associated with agonists of these opioid receptors, especially the n-opioid receptor agonists [8]. Noteworthy is that buprenorphine now is convincingly shown not to have an analgesic ceiling effect and it is not an antagonist to other μ-agonists, but rather has supra-additive analgesic effects when administered with pure μ-agonists, such as morphine or oxycodone [8]. There is a ceiling phenomenon on the respiratory effect of buprenorphine and it appears to cause less gastrointestinal dysfunction than morphine in equianalgesic doses [8].

Fig. 1

Diagram of study design and time-line.

Buprenorphine is highly lipophilic and therefore suitable for transdermal administration. The buprenorphine transdermal delivery system studied in this trial is designed for continuous 7 days wear at three strengths, giving a slow onset and slow offset with about 6 days of stable delivery of 5 μg, 10 μg, or 20 μg buprenorphine per hour. With the most recent morphine equivalent estimates of 25-100 [8], the 10 μg/h-patch delivers buprenorphine approximately equal to a morphine dose of 0.25-1.0 mg/h, or 6-24 mg per 24 hours.

Following removal of the patch, concentrations decrease to about one-half in 12 hours, and then decline more gradually with an apparent half-life of about 26 hours. These pharmacokinetics should minimise some of the adverse effects typical of fast onset, fast offset delivery of potent opioids. It allows a practical means of administering the drug to patients who may be unable to take medication orally and to elderly patients with cognitive dysfunction or impaired memory.

Purdue Pharma L.P. of USA has developed this buprenorphine transdermal system, registered as Norspa^n® in the Nordic countries and BuTrans^® in the UK. Steady-state plasma concentrations are reached by day 3 of exposure to Norspan^® / BuTrans^®. No accumulation of buprenorphine plasma concentrations have been observed in subjects exposed to Norspan^® / BuTrans^® for over 60 days.

2.2 Patients: recruitment, inclusion and exclusion criteria

Patients with osteoarthritis pain were recruited to this randomised, double-blind, placebo-controlled, parallel-group, multicentre study at the 19 investigator centres in Denmark, Finland, Norway, and Sweden. Patients were recruited mostly from pain clinics and rheumatology clinics and by newspaper advertisements.

2.3 Study design

This was a 6 months (24 weeks; 168 days), randomised, double-blind, placebo-controlled, parallel-group, multicentre study of the effects on osteoarthritis pain, functions, quality of sleep, and aspects of health-related quality of life with low doses transdermal buprenorphine patch or placebo patch, that was added to the patients’ NSAID or coxib regimen (Table 2; Fig. 1).

The trial consisted of two phases: the screening phase and the double-blind phase (Fig. 1). The study comprised up to 16 visits, where six visits could be carried out by telephone. The screening visit (visit 1) was conducted 5 to 9 days prior to the baseline visit (visit 2). In the double-blind phase, subjects were randomly allocated to either the buprenorphine or the placebo patch group in a 1:1 ratio. This phase of the study lasted 168 days during which the assessment visits were conducted at weekly intervals for the first 2 weeks (visits 3 and 4) and thereafter at biweekly visits, or by telephone contacts until visit 15. The safety follow up contact (visit 16), in the form of a telephone interview, took place 30 days following either discontinuation of the patient’s participation in the trial, or following completion of the study.

The enrolled patients began the study on the lowest dose of the buprenorphine, 5 μg/h, or placebo patch, whilst maintaining their NSAID or coxib analgesic regimen at a stable frequency and dose. Titration of dose of the study medication by the investigator was permitted for the entire duration of the double-blind phase, taking place in a stepwise fashion. Up-titration up to a maximum dose of 20 μg /h could be carried out only after a minimum of three days treatment of any given dose of buprenorphine patch or placebo patch.

Rescue analgesic was provided as paracetamol (acetaminophen) tablets 0.5 g for breakthrough osteoarthritis pain until the end of the double-blind phase, up to 4 g allowed daily. Every evening they recorded their use of rescue analgesic drug, i.e. number of paracetamol tablets used during that day in the patient’s diary. At the same time they recorded their pain on movement during that day in patient’s diary, using a numeric rating scale (NRS-11see below).

2.4 Screening phase

In the screening phase, the osteoarthritis pain of the patient was characterised, and compliance and tolerance of their current analgesic regimen were confirmed. The screening phase consisted of the screening and baseline visits, taking place approximately one week apart. The criteria for inclusion or exclusion, medical history, concomitant medication, vital signs, demography, and laboratory tests were checked for each patient. Informed consent was obtained, and all subjects agreed to discontinue all intermittent weak opioid and/or non-opioid analgesic therapy other than their NSAID or coxib regimen, and aspirin taken for cardiovascular indications.

2.5 Randomisation

Enrolled patients were randomised to one of the two treatment groups. Randomisation was performed using a validated computer system that automates the random assignment of subjects to randomisation numbers. Treatment allocation was in balanced, randomly permuted blocks, block size was unknown to investigators. The randomisation scheme was reviewed by the Biostatistics and Statistical Programming Department at the Sponsor’s site, and locked after approval. All patients, investigators, and study centre and Sponsor personnel were blinded to the medication codes. The randomisation schedule was filed in a secure location in a manner such that blinding was properly maintained throughout the study.

2.6 Double-blind phase

The buprenorphine and placebo transdermal patches were administered in 5, 10, and 20 μg/h doses. They were identical in appearance, packed in a labelled foil pouch, containing coded treatment group identification. The medication codes were not available until the completion of the study and clinical database lock, except in case of emergency. Blinding was not broken during the study.

Patients started with buprenorphine 5 μg/h or identically appearing placebo patch, titrating up to the 10 μg/h and 20 μg/h as needed. Reduction of dose was also allowed if adverse effects and analgesic effect indicated this.

2.7 Rescue medication

The rescue medication supplied was paracetamol (acetaminophen) 0.5 g tablets taken every 4-6 hours as needed; with a daily maximum recommended dose of 4 g. Subjects had to abstain from taking paracetamol rescue for at least 24 hours prior to each study visit.

2.8 Assessments of efficacy outcomes

The study focused on one OA pain site, either the hip or the knee, defined as the primary osteoarthritis site, the joint most affected by OA pain, for each participant.

Fig. 2

The Western Ontario and McMaster Universities (WOMAC) osteoarthritis index for pain.

2.8.2 Secondary outcome variables

The secondary efficacy variables were

• the changes from baseline to the end of the double-blind period in

  1. the WOMAC Osteoarthritis Index subscale score for Stiffness (Fig. 3),

  2. the WOMAC Osteoarthritis Index subscale score for Functional ability (Fig. 4a, b, c),

  3. and the WOMAC Osteoarthritis Index Total score, i.e. the sum of the scores for pain, stiffness, and functional abilities.

• The score for pain intensity on movement during the day, recorded every evening from the evening before the randomisation visit (visit 2 = baseline value – Fig. 1) and throughout the double blind treatment phase. A numeric rating scale for painintensity (NRS-11) was used, where 0 = “no pain” and 10 = “pain is as bad as you can imagine”. Every evening before going to bed, starting during the screening phase, the patients recorded in their diary the NRS-11-answer to the following question: “Overall, what has your pain on movement been like today?”

• The daily rescue medication used (number of paracetamol 0.5 g tablets) was recorded in the diary every evening.

• The number of nights woken because of pain (sleep disturbance) and quality of sleep (very poor, poor, fair, good, or very good) during the last 7 nights, recorded at each visit.

• Patient’s Global Impression of Change (PGIC) in their overall status at the end of the 6 months double blind phase, or when the patient withdrew from the study. The PGIC was scored from 1 = ‘very much improved” to 7 = ‘very much worse’, i.e. the lower the scores, the better the patient felt by the end of the study period (Table 3).

Table 3

Patient’s Global Impression of Change.

Since the start of the study, my overall status is (check only one choice):
1.	very much improved
2.	much improved
3.	minimally improved
4.	no change
5.	minimally worse
6.	much worse
7.	very much worse

• The time that the patients remained in the study, i.e. from randomisation – till completing the double blind phase after 168 days, or when the patient withdrew from the treatment due to lack of effect, intolerable side effects, or any cause, which the patient did not need to explain.

Fig. 3

The Western Ontario and McMaster Universities (WOMAC) osteoarthritis index for stiffness.

2.9 Safety assessments

For the purposes of safety reporting, the study period was defined as the interval between the time of randomisation and the end of the follow-up period. Safety assessments consisted of monitoring and recording all adverse events and serious adverse events, blood and urine β-hCG pregnancy testing for women of child bearing potential, vital signs, and physical examinations. Blood samples for haematology and blood chemistry were taken at screening and at the final visit, and the investigators reported all abnormal values as adverse events if they were evaluated to be clinically significant. Adverse events were recorded when occurring immediately following study drug administration, up to 7 days after the outcome visit, or up to 30 days after for serious adverse events. All efficacy and safety assessments used are considered standard and were in compliance with both national and international regulatory guidelines.

2.10 Statistical analyses

The intention to-treat (ITT) population comprised all subjects enrolled into the study. The full analysis population was defined as all subjects who received at least one dose of the study drug and for whom at least one post-dose observation was recorded for the primary efficacy outcome measure. The safety population was defined as all subjects who had received at least one dose of treatment and had at least one safety assessment. The per-protocol (PP) population was a subset of the full analysis population and consisted of subjects who complied with the protocol, also those who, for any reason, chose to discontinue before the completion of the 168 days double blind phase. This was primarily to be able to assess tolerability in the start up phase of buprenorphine patch treatment, although including those who discontinued already after one or two weeks hardly would give a correct picture of the efficacy of optimally applied transdermal buprenorphine for chronic pain.

The null hypothesis was that there would be no difference between the buprenorphine patch and the placebo patch treatments in the mean change in the WOMAC OA Index for pain score from baseline to the end of the double blind assessment period.

Both the primary and secondary efficacy variables were evaluated using a General Linear Model (GLM) with treatment as a factor and the following as covariates: the primary site of osteoarthritis pain (hip or knee), the season (which month) in which each patient entered the study, the baseline values of the four WOMAC scores and NRS-11 scores for movement-related daytime pain, and these values at the 19 investigator centres.

The primary efficacy variable was analysed in both the ITT and PP populations. All other efficacy variables were analysed in the full ITT analysis population only.

The average daily rescue medication used during the last month of the study was summarised by treatment group for patients in the ITT population.

Fig. 4

a, b, c The Western Ontario and McMaster Universities (WOMAC) osteoarthritis index for functional abilities.

Fig. 5

European Health-related Quality of Life questionnaire.

Fig. 6

Visual Analogue Scale for general state of health.

The number of nights woken because of pain and the quality of sleep recorded for the 7-day period prior to each visit, as well as the patient’s global impression of change in their (health) state (PGIC) at study completion were summarised as ordered categorical data and analysed using the Wilcoxon two-sample test.

The duration of test drug treatment (168 days or time to withdrawal if the 6 months double blind treatment period was not completed) was analysed using proportional hazards modelling, with treatment, primary site of osteoarthritis pain, gender, adverse effects, and season as factors.

In the exploratory efficacy analysis, each question in the EuroQoL EQ-5-D questionnaire was summarised as categorical data.

The sample size was based on the primary endpoint: with a sample size of 160 subjects, 80 could be allocated to each treatment group, giving the study an 80% power at the 5% significance level in order to detect a 1.0 difference between treatment groups in the WOMAC OA Index of Pain, with a common SD = 2.25, based on a previously conducted trial (BUP.CLIN0001data with the Sponsor). Assuming an approximate 30% drop out rate, then 224 eligible subjects were required for randomisation with 112 patients assigned to each treatment group. The effect size was estimated to be approximately 0.44 (=1/2.25).

3.1 Patient recruitment and characteristics

Patients were recruited mainly from pain clinics and rheumatology clinics at the 19 investigator centres, as well as through newspaper advertising. After the protocol of this study had been finalized and approved by all appropriate ethical and regulatory authorities, early in 2004, there was an increasing awareness of the adverse effects of NSAIDs and COX-2-specific inhibitors in elderly patients, which eventually resulted in updated guidelines for pharmacological treatment of pain in elderly patients in 2008-2009 [2,3]. These guidelines now advice against prolonged use of NSAIDs; they recommend that NSAIDs should be used only for severe flareups of pain, and for brief periods only [2,3]. This created increasing difficulties in recruiting patients who were taking NSAIDs in high enough and stable doses for 6 months, as required by the inclusion criteria of this trial. The original estimates of sample size of 224 were therefore adjusted to 200 (in an Amendment to the protocol), with some loss of statistical power of the study. In the end, the drop out rate was 44%, higher than the estimated dropout rate of 30% used in estimating sample size. This also decreased the power of the trial to avoid Type I errors, i.e. concluding that there is no difference between groups, when in reality there is a difference in effects between the buprenorphine and the placebo patch.

Fig. 7

Investigators’ evaluation of possible drug abuse or drug diversion.

Of recruited and screened patients, 199 patients were included into the double-blind phase of the study from August 2004 to September 2005. Of these, 100 patients were randomised to receive the buprenorphine transdermal patch and 99 the placebo transdermal patch (Fig. 8). There were more female patients in the buprenorphine group (72%) than in the placebo group (65%) (Table 4). There were no differences between the active and placebo groups in age or weight (Table 4), in joints primarily affected by OA, and in radiographic severity grade of the OA-joints (Table 5). The primary site of osteoarthritis pain was in a hip in 73 (36.7%) and in a knee in 126 (63.3%) patients, the majority having radiographic OA severity grade II (moderate) and grade III (severe), only 8 patients in each group had grade IV (very severe) OA (Table 5).

At baseline there were no differences between the overall mean values in the buprenorphine and the placebo groups in WOMAC-OA scores for pain (Fig. 9), for abilities to perform activities of daily functions (Fig. 10), for total score (Fig. 11), and baseline NRS-11 intensity score for daytime pain on movement scored each evening for the preceding daytime (Fig. 12), WOMAC OA-score for stiffness, quality of sleep, quality of life, VAS of state of health, heart rate, and blood pressure (data for the last six variables are not shown).

The number of patients recruited varied from 6 to 15 (mean 11) at each of 18 investigator centres, but one investigator centre recruited only 1 patient. There were statistically highly significant differences between the 19 investigator centres in the patients recruited, as shown by the differences in baseline values of the main outcome variables:

Fig. 8

Patients screened and included, randomised, completers, withdrawals and reasons for early discontinuations.

• WOMAC OA score for pain (mean centre score for pain varied from 8.1 to 14.3; P = 0.0035),

• WOMAC OA score for functional abilities (mean scores from 30.3 to 50.5; P = 0.0003),

• and WOMAC OA total score (centre mean from 41.9 to 69.5; P = 0.0028).

However, the NRS-11scores for movement-related pain during daytime at baseline did not differ significantly between centres (daytime movement-related pain intensity, centre means varied from 3.7 to 6.1 (P = 0.089)).

3.1.2 The per protocol population (PP) and sensitivity population

The PP analysis comprised 175 patients: 85 from the buprenorphine patch and 90 from the placebo patch group (Table 6). A sensitivity population was also included in the analyses for the primary outcome measure, comprising 186 patients from the ITT population, because there were 13 patients (7% of all participants) randomised and given first dose of study drug who should have been excluded at screening because they violated at least one of the inclusion or exclusion criteria. This sensitivity analysis demonstrated that the results were the same for the ITT-analyses with or without these 13 patients (data not shown).

Fig. 9

WOMAC Osteoarthritis Index score for pain from baseline (visit 2) to end of treatment (EOT) at 6 months (visit 15) or earlier withdrawal.

Fig. 10

WOMAC Osteoarthritis Index for abilities to perform activities of daily living from baseline (visit 2) to end of treatment (EOT) at 6 months (visit 15) or earlier withdrawal.

Table 6

Patients randomized, withdrawals, completers, intention to treat (ITT), per-protocol (PP), and safety analysis set.

	Treatment N (%)		All
	7-day buprenorphine patch	Placebo patch
Randomized subjects	100 (50.3)	99 (49.7)	199(100.0)
Exposed subjects	100 (50.3)	99 (49.7)	199(100.0)
Safety analysis set	100 (50.3)	99 (49.7)	199(100.0)
Withdrawals	54(61.4)	34 (38.6)	88 (100.0)
Completers	46(41.4)	65 (58.6)	111 (100.0)
1TT population	100 (50.3)	99 (49.7)	199(100.0)
PP population	85 (48.6)	90 (51.4)	175(100.0)
Screened subjects	100 (50.3)	99 (49.7)	199(100.0)

3.1.3 Withdrawals before completion of the 6 months double blind treatment phase

As many as 88 patients (Fig. 8; Table 6) discontinued prematurely; adverse events caused 31 to withdraw in the buprenorphine group, 2 in the placebo group. Lack of efficacy caused 7 patients to withdraw in the buprenorphine group, 12 in the placebo group. Four patients in the buprenorphine group and 5 in the placebo group withdrew without giving a reason. Protocol violations with invalid data occurred in 5 patients in the buprenorphine group and in 6 patients in the placebo group. The Kaplan-Meier survival plot shows similar dropout curves for the two groups (Fig. 13); a slightly higher risk of early withdrawal in the buprenorphine patch group than in the placebo group was not statistically significant. The risk of withdrawal was not significantly associated with gender, adverse events, or baseline WOMAC OA index for pain (data not shown).

Fig. 11

WOMAC Osteoarthritis Index total score (for pain, activities and stiffness) from baseline (visit 2) to end of treatment (EOT) at 6 months (visit 15) or earlier withdrawal.

Fig. 12

Movement-related daytime pain score every evening on a NRS-11 pain intensity scale from baseline (visit 2) to end of treatment (EOT) at 6 months (visit 15) or earlier withdrawal.

Thus, only 46 patients in the buprenorphine group and 65 patients in the placebo group (total 111, 56% of all included patients) completed the entire 168 days double-blind treatment phase of the study.

The duration of patch treatment, i.e. how long the patients remained in the double-blind phase until withdrawing, was an important safety and tolerability outcome measure. Therefore, their premature withdrawal was not by itself a protocol violation, and those with valid data are included in the analyses of the PP-population (Fig. 8; Table 6), although it may give a biased, negative picture of efficacy of the buprenorphine patch when the very early quitters are included (see above).

Table 7

Transdermal dose of buprenorphine (μg per hour) for double-blind period.

	Treatment
	7-day buprenorphine patch	Placebo Patch
Analysis set
1TT population	100(50.25%)	99 (49.75%)
Dose (μ/h)
Mean (Std)	11.0 (5.7)	13.6 (6.4)
Median	10.0	10.0
[Min;Max]	[5.0; 20.0]	[5.0; 20.0]
Mean number of titrations per month
Mean (Std)	1.6 (1.4)	1.1 (1.0)
Median	1.1	0.7
[Min;Max]	[0.3; 7.5]	[0.3; 4.6]

3.2 Efficacy

The primary efficacy variable was normally distributed both in the ITT-population and the PP-population (data not shown).

3.2.1 Primary efficacy variable

The primary outcome measure, the change in the WOMAC OA score for pain, i.e. the 5-verbal categories of pain-intensities at five different situations (pain while walking on flat surface, pain when walking up or down stairs, pain that disturbs sleep, pain while sitting or lying down, pain while standing (Fig. 2)) from baseline to the end of the 6 months double-blind treatment phase in the ITT population showed a clear tendency for greater reduction in pain in the buprenorphine group (from 10.8 to 7.5 = - 3.2) compared with the placebo group (10.6 to 8.3 = - 2.3) (Fig. 9; Table 8). In the PP population, that comprised completers and withdrawers, 25 of whom withdrew early (Fig. 9), before optimal up-titration of dose, the changes were from baseline 11.0 to EOT 7.5 in the buprenorphine group and from 10.7 to 8.3 in the placebo group. (Table 9). The differences at the end of treatment, however, did not quite reach statistical significance in the ITT-population (treatment difference: -0.86; 95% CI [−1.76; 0.04]; P = 0.061), or in the PP population (treatment difference: -0.96; 95% CI [-1.95; 0.03]; P = 0.058).

The effects on hip-and knee-pain were similar (Tables 8 and 9).

Fig. 13

Kaplan-Meier plot of fraction of patients remaining until 168 days, illustrating withdrawal of 54 patients in the buprenorphine patch group and 34 patients in the placebo patch group before end of double blind treatment phase. Completers are not included in this Kaplan-Meier plot.

There was no significant effect of the seasons (month) in which patients entered the study (data not shown).

3.2.2 Subgroup analysis of primary outcome in patients with moderate and severe radiographic osteoarthritis (OA grade II and III), excluding 16 patients with very severe osteoarthritis (OA grade IV)

The rational for this subgroup analysis is that those with the most severe osteoarthritis, with OA grade IV – very severe radiographic osteoarthritis, cannot be expected to have much effect of this relatively low dose buprenorphine, equivalent to approximately 0.2-1 mg of morphine per hour administered parenterally [8].

Analysis of the primary outcome measure for patients with grade II or grade III osteoarthritis severity (Tables 5 and 10) did show a statistically significant difference between the buprenorphine and the placebo patches (treatment difference: -1.00; 95% CI [-1.94; −0.05]; P = 0.039).

Table 9

WOMAC OA Index of Pain, per protocol(PP)-analysis.

Mean (SD)	Treatment		All
	Active	Placebo
Analysis set
PP -population	85 (48.57%)	90 (51.43%)	175(100.00%)
Hip pain
N	35	31	66
Baseline	11.0(2.3)	10.6 (3.3)	10.8 (2.8)
Change	–3.3 (3.8)	–1.8 (4.1)	–2.6 (4.0)
EOT	7.7 (3.9)	8.9 (3.6)	8.3 (3.8)
Knee pain
N	50	59	109
Baseline	10.9 (2.6)	10.7(2.7)	10.8 (2.7)
Change	–3.5 (4.0)	–2.7 (3.7)	–3.1 (3.8)
EOT	7.4 (3.4)	8.0 (3.6)	7.7 (3.5)
All
N	85	90	175
Baseline	11.0(2.5)	10.7 (2.9)	10.8 (2.7)
Change	–3.4 (3.9)	–2.4 (3.9)	–2.9 (3.9)
EOT	7.5 (3.6)	8.3 (3.6)	7.9 (3.6)
Difference between means of buprenorphine and placebo	–0.96	95%C1 [–1.95; 0.03]	P = 0.0582
Difference between means of knee and hip	0.09	95%C1 [–1.05; 1.22]	P = 0.8797

Table 10

Subgroup analysis of WOMAC OA Index of Pain of patients from ITT population of Table 8 with radiographic grad II (moderate) and grade III (severe).

	EOT pain	SD	95%C1	P-value
Active buprenorphine patch (N =85–8 = 77)	7.21	0.40
Placebo patch (N = 90–8 = 82)	8.21	0.38
Difference between means of buprenorphine and placebo	–1.00		[–1.94; –0.05]	0.0385

1. In each treatment group, 8 patients with grade IV (very severe) osteoarthritis were excluded in this subgroup analysis. Study treatments were assumed to be insufficient for patients with very severe OA.

3.2.3 Exploratory endpoint

3.2.3.1 The EQ-5-D quality of life questionnaire (Fig. 5)

This variable showed an overall improvement in all areas at the end of the study in both treatment groups. Collectively, there was an increase in the number of patients reporting no problems, and a decrease in patients who at the end of the trial answered either ‘some’ or ‘severe’ problems for the individual questions. A tendency for more improvement was seen in the buprenorphine group compared with the placebo group for the subscale of mobility, but no difference was seen between the groups in the subscales for self care, usual activities, pain and discomfort, anxiety and depression (data not shown).

Table 16

Patients general impression of change of status at end of double blind observation period. ITT population.

N (%)	Very much impoved	Much improved	Minimally improved	No change	Minimally worse	Much worse
All
Placebo	7 (7.4%)	12(12.6%)	27 (28.4%)	38 (40.0%)	9 (9.5%)	2 (2.1%)
Active	15(16.7%)	21 (23.3%) 33 (17.8%)	22 (24.4%)	19 (21.1%)	12 (13.3%)	1 (1.1%)
All	22 (11.9%)	33 (17.8%)	49 (26.5%)	57 (30.8%)	21 (11.4%)	3 (1.6%)
Buprenorphine patch mean score Placebo patch mean score					2.9444 3.3789	P = 0.0171

Fig. 14

Patient’s Global Impression of Change (PGIC) in overall status at end of treatment.

3.2.3.2 The mean Visual Analogue Scale (VAS) for state of health

At baseline, at the end of the double-blind phase, and the change during the study period, this variable, although improving somewhat with time, did not showed any significant differences between the two groups (data not shown).

3.3 Safety and tolerability

3.3.2 Adverse events

3.3.2.1 Overall adverse events

Application site reactions occurred in 35 patients in the buprenorphine patch group and 30 patients in the placebo patch group. The majority of the local side reactions were mild to moderate (Table 17).

There were more adverse events in the buprenorphine patch group, especially nausea, vomiting, constipation, dizziness and somnolence (Table 17). There were also more severe adverse events in the buprenorphine patch group than in the placebo patch group (Table 18).

The frequency of nausea and vomiting reported for the time intervals in days from start of double blind treatment are shown in Fig. 15. Clearly, these were more frequent in the beginning of treatment and higher in the buprenorphine group than in the placebo patch group.

Women less than 63 years generally tolerated the buprenorphine patch better than both older women and men irrespective of age (data not shown).

3.3.2.2 Serious adverse events

There were 10 serious adverse events reported: four from the placebo, five from the active treatment group, and one for a patient during screening who withdrew from the study prior to randomisation and start of treatment. One serious adverse event was considered possibly drug related, involving a patient reporting vomiting and dehydration during buprenorphine patch treatment.

Table 17

Most common adverse effects (AE).

	Treatment
	7-day buprenorphine patch			Placebo Patch
	N	%	E[*]	N	%	E[*]
ITT-population	100	100.0		99	100.0
All repoting AE	92	92.0%	589	73	73.7	258
Gastro-Intestinal disorders	57	57.0	162	25	25.3	38
Abdominal discomfort	1	1.0	1	0	0	0
Constipation	24	24.0	35	5	5.1	5
Nausea	37	37.0	69	10	10.1	11
Vomiting	16	16.0	19	2	2.0	2
General disorders and administration site conditions	61	61.0	177	40	40.4	69
Application site dermatitis	10	10.0	14	5	5.1	8
Application site eczema	9	9.0	16	5	5.1	8
Application site erythema	3	3.0	7	4	4.0	7
Application site irritation	2	2.0	2	1	1.0	1
Application site pruritis	11	11.0	34	15	15.2	17
Nervous System Disorders	45	45.0	97	18	18.2	33
Dizziness	25	25.0	45	9	9.1	11
Headache	7	7.0	9	6	6.1	14
Somnolence	4	4.0	7	0	0	0

Another serious adverse event during buprenorphine patch treatment deemed probably related to the study drug, involved abdominal pain, nausea, migraine, and constipation. This subject continued with the study. All 10 patients made complete recovery.

3.3.2.3 Probably significant adverse events

Two clinically significant abnormalities in the buprenorphine group occurred in one patient who had increased levels of blood alkaline phosphatase, deemed possibly drug related; another patient had elevated blood alanine-aminotransferase levels, deemed by the investigator as probably drug related.

Fig. 15

Frequency of nausea and vomiting reported by patients from start of double blind treatment in the time intervals shown: Days 0-20; 21-40; 41-60; 61-80; 81-100; 101-120; 121-140; 141-168 (end of double blind period).

Table 18

Adverse events by severity. ITT-population.

	Treatment
	Buprenorphine		Placebo
	N (%)	E[*]	N (%)	E[*]
Reported AEs	172(100.0%)		111 (100.0%)
Adverse Events
Severe	29 (16.9%)	25	11 (9.9%)	17
Moderate	70 (40.7%)	74	42 (37.8%)	37
Mild	73 (42.4%)	69	58 (52.3%)	41

4.1 Prolonged analgesic effect was documented for low dose transdermal buprenorphine added to NSAID or coxib treatment in opioid-naïve osteoarthritis patients

Although the chosen primary outcome variable, the WOMAC OA Index of Pain did not quite reach statistical significance (P < 0.06), in a subgroup analysis excluding the 8 patients in each group who had very severe radiographic OA, the buprenorphine patch was significantly superior to the placebo patch (P < 0.04). The trial confirmed that, in spite of a 30% withdrawal rate before 6 months, the buprenorphine patch at a median dose of 10 μg per hour effectively relieves daytime movement-related pain (P < 0.03). Even more importantly, the patients’ overall impression of change in their overall status at the end of the 6 months double blind treatment period, was significantly better in the buprenorphine treated patients than in the patients treated with the placebo-patch (P < 0.02).

The patients were allowed paracetamol ad libitum up to 4 g a day. The median consumption of paracetamol as rescue analgesic was 1 g per day and not different between the two groups. Paracetamol has a potent additive analgesic effect with NSAIDs for acute pain [16,17]. Thus, the paracetamol rescue medication may have reduced the difference between a NSAID regimen with buprenorphine patch and placebo patch.

4.2 Effects on functional abilities, quality of sleep, and quality of life: Patient’s Global Impression of Change

With better pain relief, it would be reasonable to expect that the osteoarthritis patients would be able to function better in every day life, sleep better and enjoy improved state of health and health related quality of life. With the WOMAC OA index of Functional abilities, evaluating abilities to perform 17 different activities of everyday life, an effect was indicated by the buprenorphine patch (P < 0.06). The patients with primarily hip OA pain were awakened less frequently, although not statistically significantly so (P = 0.09), in the buprenorphine group. Although, less pain may lead to more activity, more pain may be provoked by using weight-bearing OA joints, which again may mean less ability to function for a period. Such variations may become less obvious in the 24 hours-mean scores required by the patients when answering the WOMAC questionnaires (Fig. 4a-c). The movement-related daytime pain evaluated by the patient every evening clearly demonstrated better pain relief from the buprenorphine patch during movement compared with the placebo patch (P < 0.03 – Table 14, Fig. 12).

The subgroup analysis of the ITT-population without the 16 patients who had very severe radiographic OA, also documented a significantly better pain relief (P < 0.04; Tables 5 and 10) with the buprenorphine patch compared with the placebo patch. The rational for doing this subgroup analysis was that patients with very severe OA, if they did have pain relief at all, and then were more active, for sure would have a rebound of their pain intensity, both at rest and during movement.

The measures of quality of sleep and quality of life, as well as general health state improved somewhat in both groups, especially during the first few weeks of the double blind observation period. During almost 7.5 months we had close contact with our trial patients; this is bound to have a significant context sensitive therapeutic effect, an observation that has been made repeatedly in pain trials [18]. This underscores the importance of having a watertight double blind study design [18,19].

Still, the patients’ evaluation of their Global Impression of Change at the end of the double blind treatment period significantly (P < 0.02) documented that the buprenorphine patch resulted in an improved overall situation for the osteoarthritis patients with pain (between “minimally improved” and “much improved”) compared with the placebo treated patients (mostly between “no change” and “minimally improved”) (Table 16; Fig. 14).

4.3 Safety of the 7- day buprenorphine patch

There were equal numbers patients with serious adverse events in the two treatment groups, 10 in all. In the buprenorphine group two patients suffered serious gastrointestinal adverse events, deemed probably drug related, with nausea, vomiting, dehydration, abdominal pain. One patient continued while one discontinued the trial. Two had possibly significant elevation of liver enzymes, deemed probably related to study drug, however, paracetamol as rescue analgesic (varied from 0 to 3.5 g per day) is a more likely explanation than the buprenorphine [20].

Except for the two clearly opioid-induced gastrointestinal side effects, none of the other serious adverse events seem relevant for the safety of buprenorphine. Buprenorphine has been in clinical use for more than 3 decades without liver damage as a recognized side effect.

4.4 Tolerability of the 7 day buprenorphine patch

Skin reactions to the patch were similar in the two patch groups with about 30% of patients experiencing local irritation-type reactions, mostly of mild severity.

The tolerability of the buprenorphine patch was otherwise good, although more patients discontinued the buprenorphine patch due to side effects before the end of the 6 months double blind period than patients having the placebo-patch. These adverse effects were mostly typical opioid side effects in opioid-naïve patients, such as nausea, vomiting and constipation, less common were dizziness and sedation (Table 17). Such opioid side effects can be reduced by a more gradual increase in buprenorphine dose than was mandated by the study protocol in the present trial.

4.5 Balance between benefits and side effects

Our present study thus confirms a solid clinical observation, that not all patients are satisfied with even low dose opioid treatment of their chronic pain. This is sometimes due to lack of perceived effect on the burden of the pain condition, but more often it is due to bothersome side effects [5]. The gastrointestinal dysfunctions that are almost obligatory for opioid receptor agonists [21], clearly also occur with buprenorphine. While attention to diet and prescription of appropriate laxative medication can mitigate constipation, the upper gastrointestinal side effects (nausea, vomiting, gastro-oesophageal reflux, dyspepsia, abdominal bloating) can be reduced only by reducing the dose or discontinuing the opioid.

An alternative which is now becoming available is to administer a μ-opioid antagonist that has only peripheral effects, such as methylnaltrexone or alvimopan. These antagonists do not cross the blood-brain-barrier [21]. Naloxone, in oral controlled release depot tablets containing oxycodone, will act on the opioid receptors of the intestinal autonomic nervous system, before being transported to the liver where naloxone is almost completely transformed to inactive metabolites, so that only a very small amount of unmetabolized naloxone reaches the systemic circulation and the CNS [21].

4.6 Limitations of the trial

This study confirmed that a multicentre study risks wide variation in characteristics of the included patients, thereby decreasing sensitivity of the trial: There was a significant variation in the baseline values of most of the outcome variables between the 19 investigator centres. Pain-trial-sensitivity is always better if pain-variables, such as pain intensity and reduced function and pain-related quality of life, are appropriately severe and uniform at baseline. This is logical and intuitively obvious: The more pain and pain problems there are to relieve, the better will the trial be able to distinguish between a more potent and a less potent analgesic therapy [18,19]. When pain is only mild or moderate, or baseline pain varies a lot between patients, even morphine will appear to be no more effective than a very weak analgesic, e.g. tramadol [18].

In the present study patients were included if they had at least moderate pain in spite of their ongoing NSAID or coxib regimen. Thus patients in one investigator centres where mean (of 10 patients) baseline pain was only 3.7 on a NRS-11 scale, clearly had patients with pain at randomisation which was low enough to reduce the trial sensitivity of this study [18,19].

This study also confirms the important observation that a straight forward numeric rating scale for pain intensity during movement [22], recorded every evening by the patients, is more sensitive in measuring differences between an active analgesic and placebo than a more elaborate outcome measure, especially when the patient has to remember what a number of variables were like during 24 hours, and in his mind, estimate the mean values for this time period [23]. The various WOMAC OA subscales dependent on the patient’s memory and ability to create a 24 hour average of several aspects of their pain condition. Thus, the defined primary outcome in this study, the WOMAC OA Index of pain, where the patient has to remember what the pain was like in 5 different situations during the last 24 hours, was clearly less sensitive than the secondary outcome measure, the daytime-movement-related pain-intensity on a 11-point numeric rating scale from 0 = no pain to 10 = worst pain imaginable. The patients scored their day-time movement related pain every evening, when they were likely to remember well what the day was like.

An aspect of our trial design, which now can be criticized, was the choice of elderly OA-patients taking continuously relatively high doses of NSAIDs or coxibs as a major inclusion criterion. This would not be ethically acceptable only a few years after the study was planned and approved by the appropriate ethical and regulatory control authorities. NSAIDs and coxibes are now not recommended for continuous use in elderly patients in general, and for OA-patients only for very brief periods of episodic flare-up OA-pain [2,3].

It should not be clinical practice to initiate long term opioid treatment for chronic noncancer pain, unless the patient is documented to have a pain condition that responds to opioid analgesics. This was not performed in the present trial, and can be criticised. However, osteoarthritis pain has been considered to be mainly nociceptive type pain, which normally responds to opioid analgesics. We now know that even in osteoarthritis, much of the pain is due to neuropathic type mechanisms [24]. Neuropathic pain does not always respond to even potent opioids [25].

The post-hoc analysis where the patients with very sever OA were excluded, indicated that their pain condition was too severe for the low dose buprenorphine patch to be helpful. Thus, we probably included patients with too little baseline pain and patients with much too severe baseline pain for the analgesic drug we tested to be appropriate analgesic treatment (see paragraph 4.8).

Finally, our study can be criticised for not documenting well enough who and how many of the included patients had, intermittently, been using analgesic tablets containing codeine or tramadol before the study. Thus, an unknown number of our patients may not have been truly opioid-naïve patients. Buprenorphine, especially in the low doses used in the present study, would be expected to have less analgesic effect if the patients in fact had been using codeine or tramadol, even if only intermittently, for prolonged periods before entering the study. The so-called “weak” opioids, codeine and tramadol are well documented to cause development of opioid tolerance. Patients who have been treated with fast onset, codeinecontaining analgesics, when prescribed the low-dose buprenorphine patch, mostly continued using their codeine containing tablets alongside the new treatment with buprenorphine patch [26].

A recent review of the properties of buprenorphine emphasizes that this potent, full μ-opioid agonist has additive or supra-additive analgesic effects with other μ-agonists, that buprenorphine differs from other opioids in common use in having a pronounced antihyperalgesic effect, favourable safety profile in elderly patients and those with renal impairment, and its lack of effect on sex hormones and the immune system [27].

4.7 Strong aspects of this study

To the best of our knowledge, and according to the 2010 Cochrane review [6], this 6-months double blind, placebo controlled study is the only one published study lasting as long as half-a-year. It is therefore unique in documenting that the opioid analgesic effect is maintained for at least 6 months. It clearly also documents, that in spite of the low doses of buprenorphine, a third of the patients do not tolerate the gastrointestinal adverse effects, and therefore discontinue the treatment after a few months. A more careful, slow escalation to an effective buprenorphine dose may increase tolerability.

We selected patients carefully for any indication in their history of increased risk of problematic opioid use, and we examined the patients carefully for signs or symptoms of drug abuse or drug diversion after the 6 months study period. None were observed in any of the 100 patients randomized to prolonged buprenorphine treatment.

4.8 What would be an ideal design for a follow-up study of buprenorphine patch 5-35 μg/h for OA-pain

Too low pain intensity causes loss of assay sensitivity. Too high pain intensity causes early withdrawal. Therefore, select patients who, without analgesic treatment, have movement-related pain intensity about 6-8 on a NRS-11scale and most days have 5 or higher pain intensity at rest

• Include only truly opioid-naïve patients (also codeine and tramadol induce opioid tolerance).

• Include only patients who have opioid-sensitive pain, pain that responds to buprenorphine in the dose range to be tested. This will cause an enriched sample, but will give a true picture of efficacy and tolerability of the drug.

• Start with a low dose, and escalate slowly to analgesic effect, or to intolerable side effects occur.

• Prevent, or treat promptly side effect: Laxatives for constipation. Antiemetics for nausea.

• Offer paracetamol 1g as rescue analgesic, up to 4 g per day. And monitor liver enzymes.

• Offer an effective treatment for occasional break-through-pain: a fast onset-potent NSAID, if tolerated (e.g. ketorolac).