Oxycodone for Cancer-Related Pain: Meta-analysis of Randomized Controlled Trials

To evaluate the efficacy and tolerability of oxycodone in cancer-related pain, we conducted a systematic review of randomized controlled trials. Four studies, comparing oral oxycodone with either oral morphine (n = 3) or oral hydromorphone (n = 1), were suitable for meta-analysis. Standardized mean differences in pain scores comparing oxycodone with control groups were pooled using random-effects models. Overall, there was no evidence that mean pain scores differed between oxycodone and control drugs (pooled standardized mean difference, 0.04; 95% confidence interval [CI], −0.29 to 0.36; P = .8; I² = 62%). In meta-regression analyses, pain scores were higher for oxycodone compared with morphine (0.20; 95% CI, −0.04 to 0.44) and lower compared with hydromorphone (−0.36; 95% CI, −0.71 to 0.00), although these effect sizes were small. The efficacy and tolerability of oxycodone are similar to morphine, supporting its use as an opioid for cancer-related pain.

Approximately half of all patients with advanced cancer experience moderate to severe pain.¹ Opioids are the mainstay of treatment,² and morphine is the opioid of choice.³ When morphine is used appropriately, about 80% of patients will achieve adequate pain relief.⁴ Approximately 20%, however, will not and may need to switch to an alternative opioid.⁵ In a proportion of these patients, this is because of intolerable adverse effects associated with morphine.

Oxycodone is a semisynthetic derivative of morphine. It has been in clinical use since 1917,⁶ but patterns of use have differed worldwide, perhaps reflecting a lack of clinical studies investigating its efficacy. Historically, it was most commonly used in the United States as an opioid for mild to moderate pain in low-dose combinations with acetaminophen (paracetamol) or aspirin. In such preparations, however, its use was limited, as its dose could not be increased because of potential acetaminophen or aspirin toxicity. It was also used for moderate to severe pain in Finland (mostly by parenteral administration).

Studies conducted since 1990 have suggested that, when used in single-entity formulations and with dose titration, oxycodone is as effective as morphine.^7,8 The 1996 European Association for Palliative Care guidelines recommended oxycodone as an alternative to morphine.³ Although there has been speculation that oxycodone may have a better adverse effect profile compared with morphine,⁹ there are limited data on cancer-related pain. Since 1996, oxycodone has been relaunched in different formulations and dose strengths and in modified-release preparations, increasing its potential for use in chronic cancer pain. The success of this relaunch is indicated by English Department of Health statistics,¹⁰ which show that the percentage of annual growth in items of oxycodone prescribed from 2002 to 2003 was 43%, compared with 8% for all opioid analgesics. Annual consumption of oxycodone has increased 42-fold in the United Kingdom and 3-fold in the United States from 1999 to 2003¹¹ (Figure 1). We conducted a systematic review of the available evidence to determine the efficacy and tolerability of oxycodone for cancer-related pain.

We included randomized controlled trials comparing oxycodone with placebo or an active analgesic drug in patients with cancer-related pain. All routes of drug administration and all formulations of oxycodone were considered. Studies of combination oxycodone preparations (eg, oxycodone and acetaminophen) were excluded.

We searched the following electronic databases using a detailed search strategy for each: Cochrane Pain, Palliative and Supportive Care Register 2002; Cochrane Controlled Trials Register 2002: Cochrane Library current issue; MEDLINE (1966 to May 2002); EMBASE (1980 to May 2002); CancerLit (1960 to May 2002); CINAHL (1982 to May 2002); dissertation abstracts (2002); and SIGLE (2002). We searched reference lists of retrieved articles and other relevant literature such as pain or opioid reviews. We wrote to the manufacturers of oxycodone preparations (Napp Pharmaceuticals, Cambridge, England, and Purdue Pharma, Stamford, Conn), known oxycodone investigators, and subscribers of Palliative Medicine and selected pain journals with a request for data from unpublished trials or information about other trials we had not identified. The search strategy was repeated in April 2005, with no new studies identified. Search terms are listed in Figure 2.

The full-text versions of potentially eligible articles were obtained and independently assessed by 2 of the investigators (C.M.R. and A.N.D.). We identified duplicate publications by reviewing study name, authors, location, study population, dates, and study design and by confirming with the study authors and Napp Pharmaceuticals (the manufacturers of oxycodone in England) that each of the included reports were indeed separate studies. We identified replication of efficacy data from trials by Kalso et al¹² and Heiskanen et al¹³ contained within separate articles reporting on the pharmacokinetic outcomes from these studies. Reasons for excluding a trial were recorded.

For eligible trials, both investigators independently extracted data from the article using a specifically designed data extraction form. This form recorded the following: publication details, patient population details, nature of pain if described, interventions, outcome measures used, analgesic (efficacy) results, adverse effects, quality-of-life scores, patient preference, withdrawals, and trial quality criteria. We extracted data on reported methods of concealment of allocation¹⁴ and the blinding of therapists, patients, and outcome assessors in each trial.

Different assessment scales were used to record pain scores in the trials (Table 1). We expressed treatment effects as standardized weighted mean differences.²⁰ For the parallel group trial,¹⁹ the standardized weighted mean difference was calculated using the Glass estimator²¹ and the standard error calculated according to equation 9 of Curtin et al.²⁰ For crossover studies, standardized weighted mean differences were estimated according to equation 11 of Curtin et al,²⁰ by dividing the treatment effect by the between- plus within-subject standard deviation of the crossover differences. To estimate the standard deviation of the crossover differences, a common between-period intraclass correlation coefficient of 0.2 was estimated using individual level data available only for the trial by Heiskanen and Kalso.¹⁸ Estimates of the variance of the crossover effect sizes were then derived by equation 14 of Curtin et al.²⁰ Because the trials used different control groups (morphine or hydromorphone), and because there was evidence of between-trial heterogeneity, trials were pooled using random effects meta-analysis.²² We analyzed pain scores recorded on the final day on each study drug to ensure that steady state had been reached. Data on the presence or absence of 16 common opioid-related adverse effects were obtained from the authors of each crossover trial. We used the marginal odds ratio (OR) estimate described by Becker and Balatgas²³ to obtain ORs and their standard errors for the crossover trials. For each adverse effect, these ORs were combined with the corresponding ORs derived from the published report of the parallel group trial. The I² statistic was used to assess the extent of between-study variation in estimates,²⁴ and sources of heterogeneity were explored using meta-regression.²¹ Analyses were conducted using Stata statistical software, version 8.0²⁵ (meta-analysis of pain outcomes) and RevMan version 4.2.7 (Nordic Cochrane Centre, Copenhagen, Denmark) (meta-analysis of adverse effects).

The search strategy yielded 104 REFERENCES, of which 6 trials met inclu sion criteria (Figure 3). Of these, 1 was a single-dose study evaluating the analgesic potency and duration of action of intramuscular oxycodone against intramuscular morphine and codeine phosphate.¹⁵ Of the remaining 5 reports, 3 crossover trials compared oral oxycodone with oral morphine^7,16,18; 1 crossover trial compared oral oxycodone with oral hydromorphone¹⁷; and 1 parallel group trial compared oral oxycodone with oral morphine¹⁹ (Table 1).

None of the eligible studies reported data in a form suitable for meta-analyses, so we contacted the authors or the sponsoring drug companies for additional data. We also requested additional data on the presence or absence of adverse effects examined by the trials. Individual patient data were obtained for the study by Heiskanen and Kalso.¹⁸ For the studies by Bruera et al,¹⁶ Hagen and Babul,¹⁷ and Mucci-LoRusso et al,¹⁹ we were provided with mean within-patient differences in pain scores (comparing the first and final study days) and an estimate of the standard deviation. Analyzable data were unavailable for the studies by Kalso and Vainio⁷ (which reported no statistically significant difference in visual analog scale ratings between the morphine and oxycodone groups, but more nausea with oral morphine) and Beaver et al¹⁵ (which demonstrated that intramuscular oxycodone was 0.68 times as potent as intramuscular morphine but had a slightly shorter duration of action). Thus, 4 studies were included in the meta-analysis.

Only 1 trial (Heiskanen and Kalso¹⁸) reported on the method of concealment of allocation to treatment (the randomization code was kept by the hospital pharmacist), although information about whether restricted randomization was used in generating the allocation sequence was not available in the trial report (Table 2). All of the trials used matched placebo tablets to blind the patient and clinician. No studies indicated whether analysis by intention to treat was undertaken. In all included trials, patients who withdrew from the study for any reason were excluded from the final analyses reported herein. The numbers who withdrew from each trial were as follows: for Bruera et al,¹⁶ 32 entered and 9 withdrew; for Hagen and Babul,¹⁷ 44 entered and 13 withdrew; for Heiskanen and Kalso,¹⁸ 45 entered and 18 withdrew; and for Mucci-LoRusso et al,¹⁹ 101 entered and 22 withdrew (Table 1). None of the publications reported whether the outcome assessor was blinded to treatment. Each trial reported that patients in both treatment groups had their opioid doses titrated in a similar manner until stable doses were obtained. The trials were of short duration, lasting from 10 to 20 days.

In pooled analyses, we found no evidence that mean pain scores differed between the oxycodone and control groups (mean difference in standardized pain scores, 0.04; 95% confidence interval [CI], −0.29 to 0.36; P = .8) (Figure 4). There was evidence of heterogeneity between the study estimates (I² = 62%; heterogeneity, P = .05). The pooled standardized difference in pain scores for the 3 studies that compared oxycodone with morphine was 0.20 (95% CI, −0.04 to 0.44; I² = 0%) and the standardized difference for the study that compared oxycodone with hydromorphone was −0.36 (95% CI, −0.71 to 0.00; for difference in effect estimates, P = .1). Based on the pooled standardized differences we observed and the standard deviations observed in the individual trials, we estimate that for oxycodone compared with morphine or hydromorphone, the pooled standardized differences represent only 2 to 3 mm on a 100-mm visual analog scale. It is suggested that a clinically important difference is a change of 2 points on a 0- to 10-point pain intensity scale (equivalent to 20 mm on a visual analog scale),^26,27 indicating that the standardized differences we detected are unlikely to be clinically important or meaningful to patients.²⁷

The results of other outcomes described in the included studies are detailed in Table 1. In summary, no differences in patient preference or quality of life were demonstrated, although the study by Heiskanen and Kalso¹⁸ suggested that nighttime acceptability of morphine was better than that of oxycodone. Because different measures were used and the results were not reported in sufficient detail, they could not be combined in meta-analyses.

The point estimates for the pooled ORs comparing oxycodone with control groups were 0.75 (95% CI, 0.51-1.10) for nausea and 0.72 (95% CI, 0.49-1.06) for vomiting (Table 3). There was substantial evidence of heterogeneity in estimates of the association of oxycodone with dry mouth and drowsiness (I² = 74% and I² = 71%, respectively). When the meta-analysis was repeated using only data from the trials with morphine as the control treatment, the pooled OR favored oxycodone for dry mouth (OR, 0.56; 95% CI, 0.38-0.83) and drowsiness (OR, 0.72; 95% CI, 0.47-1.1). Overall, the discontinuation rate due to adverse events was 13% (29/222) when data from all of the studies were combined; as many as 90% of patients experienced opioid-related adverse effects in each trial (Table 4). Discontinuation rates due to adverse events were similar in the oxycodone and control groups.

This study is the first we are aware of to display the accumulated evidence to date on which current prescribing of oxycodone in cancer-related pain is based. We found no clinically important differences between the analgesic efficacy or the adverse effect profile of oxycodone compared with morphine. Although only 160 patients were included in the meta-analysis, the 95% CI for the effect of oxycodone vsmorphine is narrow and excludes any clinically important differences between these 2 drugs, in that the upper limit of the CI (standardized difference, 0.44) is consistent with a difference of only 6 mm on a 100-mm visual analog scale and the lower limit (−0.04) is consistent with a difference of 0.5 mm. For oxycodone vs hydromorphone, the equivalent figures are 0 and 6 mm, respectively. These differences are much lower than those that are suggested to be clinically meaningful (a change of 20 mm on a 100-mm visual analog scale^26,27).

Our findings highlight the paucity of data in this area, and the results have to be interpreted with caution. The study reports did not allow us to be confident about the internal validity of the trials. Only the trial by Heiskanen and Kalso¹⁸ reported an attempt to conceal treatment allocation, stating that the hospital pharmacist held the codes. This does not necessarily guarantee that concealment of allocation was successful in preventing bias, as most concealment processes can potentially be subverted²⁸ and no attempt to assess whether bias was indeed avoided (eg, using the methods of Berger and Exner²⁹) was reported in any of the included studies. Inadequate or unclear concealment of treatment allocation is associated with exaggeration of treatment effects³⁰ and is therefore unlikely to explain the absence of clinically important differences between the oxycodone and control groups found in our meta-analysis. In each of the included studies, patients who withdrew for any reason were not included in the final analyses comparing pain scores between the oxycodone and control groups (Table 1), possibly resulting in attrition bias, which might further threaten the validity of the individual studies. However, because the discontinuation rates due to adverse events were similar for the oxycodone and control groups in all studies, it seems unlikely that the potential for attrition bias explains the results of our meta-analysis.

The percentage of patients experiencing adverse effects and discontinuing treatment due to adverse events was considerable and in line with discontinuation rates from other studies of opioids in cancer and noncancer populations.^31-33 The prevalence of adverse effects was higher in studies that directly questioned patients about adverse effects compared with those that relied on self-reporting. Because the presence of adverse effects is one of the reasons patients are unwilling to continue or increase doses of medication for pain relief,³⁴ these findings reemphasize the need for active questioning about and aggressive management of opioid-related adverse effects. However, the short duration of these studies also meant that useful data about long-term adverse events such as aberrant drug-seeking behavior were not obtained. Although we would not have expected aberrant drug-seeking behavior to be a significant problem, as this adverse event appears to be rare in the cancer population,³⁵ evidence from trials might help to dispel the fear of addiction that often hampers good pain control.³⁴

The small number of studies retrieved and the short duration of the studies are perhaps a reflection of the difficulties of conducting clinical trials in this group of patients.^36,37 High attrition rates due to worsening of the underlying disease or intercurrent illness mean that investigators try to minimize losses to follow-up by designing trials of short duration. Although this approach can provide robust short-term data, it means that results on longer-term efficacy or rates of adverse effects are often not available. Short-term data on comparative effectiveness and adverse events, however, provide important information for clinicians managing cancer pain, to inform patients of the likelihood of beneficial and immediate adverse effects and to reinforce the need for appropriate management of these early adverse effects so that adequate dose titration can be achieved and satisfactory pain relief obtained.

The studies recruited outpatients from general cancer or palliative care patient populations, with mixed types of pain from a variety of cancers. Where reported, the patients were treated as outpatients, although daily contact was made. It seems likely that the patient population in the trials is representative of patients with cancer-related pain seen in usual practice (Table 1). It is unclear whether these results are of relevance to patients with chronic noncancer pain, which represents a larger group of patients for whom opioids are considered. Previous systematic reviews^32,38 have examined the efficacy of opioids in various types of noncancer pain, but only 1 review³³ has attempted to compare relative efficacies of different opioids. These reviews have confirmed the efficacy of opioids in a variety of chronic pain conditions and have demonstrated that adverse effects are as common as in cancer populations. To our knowledge, no studies comparing oxycodone with other opioids for moderate to severe pain have been conducted in patients with noncancer pain, but there is no clinical or pharmacologic reason to believe that one opioid would be more effective than another in one setting (eg, cancer patients) compared with another (eg, patients with chronic noncancer pain).

Morphine, in both normal- and modified-release formulations, has been the first-line opioid in England for the management of moderate to severe cancer pain. In this review, we did not find any important differences between oxycodone and morphine. Oxycodone is almost 4 times more expensive than morphine in England, and there is less general experience of its use. Thus, there is no reason to challenge the recommendation to use morphine as a first-line agent for cancer pain. There is a need, however, for larger trials of longer duration designed to obtain comparative efficacy and adverse event data for opioids for moderate to severe pain in cancer and noncancer populations.

Correspondence: Colette M. Reid, MB, Department of Palliative Medicine, University of Bristol, Horfield Road, Bristol, Av BS2 8ED, England (Colette.reid@bristol.ac.uk).

Accepted for Publication: November 3, 2005.

Author Contributions: Drs Reid and Martin had full access to all the data in the study and take joint responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: The Department of Palliative Medicine, University of Bristol, has received funding from and has collaborated in drug trials with Napp Pharmaceuticals.

Acknowledgment: We thank Napp Pharmaceuticals and Tarja Heiskanen, MD, PhD, for providing additional data. Frances Fairman offered advice on Cochrane methodology.

Additional Information: A fuller version of this review and its findings will be available through the Cochrane Library.