A meta-analysis of efficacy and tolerability of buprenorphine for the relief of cancer pain

This study aimed to synthesize available evidence on the analgesic efficacy of buprenorphine in treating cancer pain and related adverse effects. We searched electronic databases for randomized controlled trials, assessing the efficacy of buprenorphine, regardless of delivery system. The primary endpoints were patient-reported ‘pain intensity’ and ‘pain relief’. Statistical heterogeneity among included studies was assessed with the I2 test. The summary relative risk (RR) and 95% CI were derived, if two or more studies reported the similar outcome. Sixteen RCTs (n = 1329) with buprenorphine were included: 8 transdermal (TD), 5 sublingual (SL), 2 intramuscular injection (IM) and 1 subcutaneous infusion (SC) studies; with both SL and IM routes being assessed in one study. Only a few studies reported the same outcome in a similar way, creating difficulty for pooling of the outcome data. Many studies had a high risk of bias. In 2 studies (n = 241), the ‘global impression change’ was significantly different between TD buprenorphine and the combined placebo and morphine (RR 1.35, 95% CI 1.14-1.59; I2: 42%); the ‘number-needed-to-treat’ (NNT) was 4.9 (95% CI: 3.1-10.9). In 2 studies (n = 331), ‘requirement for rescue SL buprenorphine’ was comparable between TD buprenorphine and placebo (RR 1.25, 95% CI 0.71-2.18; I2 : 40%). In 2 studies (n = 141), ‘incidence of nausea’ was less in TD buprenorphine (RR: 0.38, 95% CI: 0.2-0.71, I2: 0%, NNT: 9.3, 5.6-28.5). Due to the small number of participants in a small number of studies, the results of the present review provide insufficient evidence to position adequately the use of buprenorphine in treatment of cancer pain. Large multicenter RCTs that compare TD buprenorphine with standard analgesic treatment is needed to position TD buprenorphine in the therapeutic armamentarium of cancer pain treatment. Electronic supplementary material The online version of this article (doi:10.1186/2193-1801-3-87) contains supplementary material, which is available to authorized users.


Introduction
Worldwide cancer is one of the leading causes of morbidity and mortality. In 2012 new cases of cancer was estimated to be 14.1 million per year (ICRC 2012) and is expected to climb to 19.3 million per year in 2025 (WHO 2011;ICRC 2012). The estimated cancer-related deaths was 8.2 million in 2012 (ICRC 2012). The majority of all cancers (56.8%) and cancer deaths (64.9%) in 2012 occurred in less developed countries (ICRC 2012). When there is local and metastatic spread of cancer, complications arise and pain is an inevitable outcome. Over the years, the use of opioids has been the mainstay for treating cancer pain (Hanks 1991).
The WHO recommended the concept of 'a threestep analgesic ladder' in the treatment of cancer pain (Levy 1996;WHO 1996). In step III of the analgesic ladder, morphine is the drug of choice for the management of moderate to severe cancer pain (WHO 1996(WHO , 2009Quigley 2004). However, published studies had reported that adequate analgesia with morphine was not achieved in 10-30% of patients with cancer pain (Cherny et al. 2001;Wiffen et al. 2003;Quigley 2004;WHO 2009). Along this line, alternatives to morphine which are novel formulations of existing drugs (Hanks et al. 2001) are available.
Buprenorphine, synthesized in the late 1960s was used as a parenteral analgesic since 1978. Buprenorphine is also available in the forms of sublingual (SL) tablets or transdermal (TD) patches. It is a partial agonist at μopioid receptors (MOR) (Yaksh and Wallace 2011), an antagonist at kappa opioid receptors (KOR) (Yaksh and Wallace 2011) and a partial agonist at opiate receptorlike receptor (ORL-1) (Lutfy and Cowan 2004).
In the treatment of opioid dependence, a Cochrane review showed that maintenance dose was less effective with buprenorphine than methadone. (Mattick et al. 2008). Thus, it is of immense value to provide evidence of the efficacy and safety of opioid analgesics for cancer pain.
Reviews which addressed the efficacy of TD buprenorphine are available for chronic treatment of moderate to severe pain (Deandrea et al. 2009;Tassinari et al. 2011) and for cancer pain (Wolff et al. 2012;Naing et al. 2013). All these reviews assessed TD buprenorphine, excluding other non-TD administration of buprenorphine. A review covering buprenorphine delivered via any route for treating cancer pain would be more informative and valuable in comparison. Taken collectively, the objectives of the present study were to synthesize available evidence on the analgesic efficacy of buprenorphine in treating cancer pain and on the drug-related adverse effects.

Literature search
We searched electronic databases of MEDLINE, EMBASE, CINAHL and the Cochrane Library up to July 2013. We also searched the reference sections of the selected studies and relevant reviews for any additional studies which were not found in the initial search. We followed the search terms used in our earlier publication (Naing et al. 2013). The search strategies were a combined search terms for the cancer and for buprenorphine. The searched term for cancer included; "cancer, human" [MeSH] OR "malignancy" OR "cancer, gastrointestinal" OR "cancer, bladder" Or "cancer, breast" OR "cancer, stomach" OR "cancer, colon" OR "cancer, prostate" OR "cancer, lung". Searched term for buprenorphine included; "buprenorphine" OR "bupre*".

Selection criteria
We included studies following the PICOS criteria; (1) (Participants, P): those patients with cancer, regardless of age, gender, type of cancer and healthcare settings; (2) (Intervention, I) studies where participants in one arm should use buprenorphine, regardless of route of administration; (3) (Comparison, C) studies which compared the efficacy of buprenorphine with placebo, other opioid analgesic or no treatment; (4) (Outcome, O) the proportion of participants with the changes in intensity of cancer pain; and (5) (Study design, S): RCT in which efficacy of buprenorphine preparations was assessed in cancer patients.
Studies were excluded if they (i) were not RCTs, (ii) did not assess pain as an outcome measure, (iv) were carried out with fewer than 10 participants, (v) were conducted with those who had pain with the absence of cancer, (vi) were assessed where pain was related to treatments (chemotherapy-induced neuropathic pain), or (vii) were assessed with pain due to surgical procedures. We also excluded studies on pharmacodynamics, pharmacoeconomics, case reports and conferences reports.

Data extraction and risk of bias assessment
Two authors independently screened all citations and retrieved article(s) which were considered eligible for the present review. The two authors independently extracted data from the studies, using a piloted-data extraction form. The following information was collected: study design, sample size, participant characteristics, cancer status, drugs and their dosing, duration of study and follow-up, analgesic outcome measures, adverse events (AE's) and serious adverse events (SAE's). We resolved any discrepancy by discussion.
The two authors independently determined quality of studies following the Cochrane risk of bias tool (Higgins and Green 2011). The domains assessed were 'random sequences generation' , 'allocation concealment' , 'blinding of participants and/or outcome assessment' and they were categorized as 'low' 'high' or 'unclear' risk of bias. Discrepancies were resolved by consensus.

Statistical analysis
The primary outcomes of interest were 'patient-reported pain intensity' and 'pain relief' as measured by validated scales such as verbal rating scales (VRS), visual-analogue scales (VAS) and numerical rating scales (NRS). The secondary outcomes were incidence of buprenorphinerelated AE and SAE. We performed meta-analysis when 2 or more individual studies were suitable for pooling on the basis of similarity. Dichotomous data were compared using a relative risk (RR) and respective 95% confidence interval (CI). We assessed statistical heterogeneity by I 2 test; a value of I 2 > 50% indicated substantial heterogeneity (Higgins and Green 2011). If there was substantial heterogeneity among studies, we used the DerSimonian and Laird random effect model when pooling data and calculated summary RR and respective 95% CI. We also reported the number-needed-to-treat (NNT) for significantly different outcome. In order to test the robustness of our results, we reanalyzed the effect estimates by excluding individual studies from the meta-analysis (leave-one-out sensitivity analysis). For an assessment of reporting bias, we visually inspected the funnel plots. However, asymmetrical funnel plot may be considered due to other possible bias such as difference in methodological quality among studies (Higgins and Green 2011).
Data entry and analyses were done using RevMan Version 5 · 2 (The Nordic Cochrane Centre, Copenhagen). The present review has been reported according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) (Moher et al. 2009) (Additional file 1). A protocol of this study is available (Naing et al. 2012). Figure 1 shows the summary of study selection process. The literature search yielded 984 citations. Thirty six that examined the efficacy of buprenorphine were potentially relevant. A total of 16 studies (n = 1329) with four different routes of administration (subcutaneous, SL, IM, TD) were identified for the present review.

Results
Characteristics of the included studies (Table 1) Buprenorphine was applied TD in 8 studies, SL in 5 studies, IM in 2 studies and subcutaneous infusion (SC) in one study. Both SL and IM routes were assessed in one study (Dini et al. 1986). Both SC and IM routes were assessed in one study (Noda et al. 1989). The majority of the included studies (10/16: 62.5%) were with the sample size less than 100 (ranges: 17-189). The comparator drugs included placebo, pentazocine, morphine, TD fentanyl and tramadol. Six studies were cross-over studies with each patient serving as his/her own control, while 9 studies were parallel-group comparisons and one study was within dose comparison. In accordance with the definition of enriched described elsewhere (Straube et al. 2008), three studies were completely enriched (Bohme and Likar 2003;Sorge and Sittl 2004;Poulain et al. 2008) and one study was partially enriched (Aurilio et al. 2009). Reported outcomes in the i analysed studies included a number of validated scales subjectively used for pain intensity and/or pain relief (e.g. VAS, VRS, NRS, questionnaires), requirement for rescue drugs, duration of pain-free sleep and presence of AE's and SAE's. As several different pain scoring systems were employed in these studies, a comprehensive comparison could not be made between all of them.

Risk of bias
Overall, many studies had high risk of bias ( Table 2). Blinding of participants or outcome assessment was done in 4 studies (Bohme and Likar 2003;Sorge and Sittl 2004;Poulain et al. 2008;Aurilio et al. 2009). Most of the studies (87.5%) had short intervention duration (≤ 15 days).

SC administration
One study (Noda et al. 1989) reported that the use of SC buprenorphine at the rate of 4 g/kg/day for 48 hour in cancer patients gave satisfactory pain relief without serious complication. However, as the findings were based on uncontrolled series of only 30 patients, it should be interpreted with caution.
AE's were not consistently reported in 3 studies with the use of IM buprenorphine.

Other manifestations
When patients were switched from one opioid to another (i.e. TD buprenorphine and TD fentanyl), the effects of analgesia and respiratory depressant of the second opioid (TD fentanyl in this review) was pronounced in some patients due to 'incomplete cross tolerance' (Levy 1996). AE's in patients in cross-over studies (Likar et al. 2007;Aurilio et al. 2009) were likely to be related to this important effect.

Discussion
The present review focused on the analgesic efficacy and tolerability of buprenorphine given by four different administration routes, SC, SL, IM or TD. We were unable to synthesize the estimations in the non-TD route of buprenorphine. This was due to the (1) small number of studies with small numbers of participants included; (2) different SL retention times (3) different eventual disposition of the SL tablet (swallowed or expectorated) (Robbie 1979;Reisfield and Wilson 2007); (4) different total doses of IM administrations or (5) different outcome measures. Overall, limitations in volume or surface area of SL space could cause drugs given via SL route to possibilities of ulceration, while the TD route is a non-invasive alternative to the oral route, particularly for stable pain states  (Reisfield and Wilson 2007). The TD buprenorphine studies have provided some indication of pain improvement over short-term studies (≼15 days follow-up). Moreover, AE's such as nausea and the requirement of prophylactic antiemetic were significantly lower in the TD buprenorphine group. Published non-Cochrane reviews have assessed TD buprenorphine in particular for non-cancer pain (Deandrea et al. 2009;Rossitto et al. 2009). In these reviews, incidences of nausea and treatment termination were significantly fewer in TD buprenorphine group than in TD fentanyl group. This is comparable to our findings, although variations in the number of studies and inclusion criteria do exist. Buprenorphine has the theoretical and potential advantage of being an opioid partial agonist with low abuse potential (Robbie 1979;Bohme and Likar 2003) and approximate equal bioavailability by the SL and TD routes (Skaer 2006). Of note is the marked inequality between the groups of TD buprenorphine and placebo in a completely enriched study (Sorge and Sittl 2004). For instance, patients in the placebo group had a better starting condition and run-in phase. Also, more patients in the placebo group had received radiotherapy shortly before the onset of the study, contributing to their better pain status. Buprenorphine, when delivered via the TD route passively diffused into the systemic circulation (Bohme and Likar 2003;Sorge and Sittl 2004), providing a slower increase in serum concentration and no peak-andtrough effects as seen with the SL route (Yaksh and Wallace 2011;Al-Tawil et al. 2013). This is the reason why in many cases SL buprenorphine is used as rescue analgesic. As a matter of fact, TD buprenorphine is noninvasive and a suitable choice for cancer pain relief even in the presence of renal disease. It also has a beneficial ceiling effect for respiratory depression. A recent pharmacokinetic study has reported that the systemic exposure to TD buprenorphine was sufficiently similar between elderly (> 75 years) and younger (50-60 years) participants (Fudala et al. 2003). However, due to the heterogeneity of the published data and small sample sizes, the current meta-analysis has thus some limitations to provide sufficient evidence.
We acknowledge some limitations of this review. If not all, most of the included studies had small sample sizes (< 100 in each arm), therefore it has inadequate power to detect significant difference. Due to the small number of participants in a small number of studies, the results of this review provide insufficient evidence to position adequately the use of buprenorphine in treatment of cancer pain. One third of studies with low risk of bias administered the TD buprenorphine (compared with placebo or other opioid analgesics) when pain levels were moderate or severe, ensuring that the studies were sensitive to detect pain-related outcomes. The preferred outcome measure '50% pain relief ' , was not reported in any included studies. Moreover, chronic pain studies (cancer pain in this case) of short duration (≲ 6 weeks) have manifested greater treatment effects than those of longer duration (≳ 8 weeks). Fourteen studies (87.5%) were conducted over short intervention duration (≲ 15 days). Furthermore, severity of pain in cancer patients could be related to the result of a progression of the underlying diseases. Many cancer patients could have secondary metastases at the time of the studies (Likar et al. 2007). It was documented that marked escalation of opioid doses may be required (i.e. by 100 times or more) for those patients with solid tumours with metastases to the spine or central nervous system (Zhang et al. 2003). However, inadequate data precluded us from doing stratified analysis according to the tumour types. Furthermore, data were based on mixed groups (cancer pain and non-cancer pain) in some studies and we could not analyse cancer patients in such studies separately. The effect of such a selection bias was considered with caution. Hence, accuracy of effect estimation in the present review remains a concern.

Conclusions
In the treatment of cancer pain, low level of evidence exists for the benefit of buprenorphine particularly with TD administration. Large multicenter RCTs that compare TD buprenorphine with standard analgesic treatment is needed to position TD buprenorphine in the therapeutic armamentarium of cancer pain treatment. As the invasive route is relatively unfavourable for use in the treatment of patients with cancer pain, there is no justification for further research with IM buprenorphine.