TABLE 2
FROM:
Controlled clinical trials in cancer pain. How controlled should they be? A qualitative systematic review
R F Bell, T Wisløff, C Eccleston and E Kalso
BACK TO ARTICLETable 2. Statistics
| Trial | Design | Primary outcome | Sample-size calculations | Type of statistical analyses | Comment on statistical analyses | General comments | |
|---|---|---|---|---|---|---|---|
| 1 | Boureau et al (1992) | CR morphine suspension vs CR morphine tablets (E) | Pain intensity | No sample-size calculations mentioned |
t-test, 
2 test, ANOVA | Tests appear appropriate | Authors conclude that CRM suspension is as effective as CRM tablets. The tests performed show only no evidence of effect, not evidence of no effect |
| 2 | Broomhead et al, 1997 | SR morphine once a day formulation vs SR morphine twice daily formulation (E) | Elapsed time to remedication/total amount of rescue medication (mg) | Sample-size calculations performed, based on results from phase one | ANOVA, Dunnett's multiple comparison procedure,
2 test, Fisher's exact test, Cochran-Mantel–Haenzel chi squared | The adjustment of the significance level due to large number of comparisons is appropriate, as are the statistical analyses | |
| 3 | Bruera et al (1996) | SR hydromorphone vs IR hydromorphone (E) | Pain intensity | No sample-size calculations are mentioned | ANOVA, Cochran-Mantel–Haenzel test | Tests appear appropriate | |
| 4 | Bruera et al (1998) | CR oxycodone vs CR morphine (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed | Three-way-ANOVA, two-way ANOVA,
2 test, Pearson correlation | Tests appear appropriate | Authors conclude that the efficacy of CR oxycodone is at least equal to CR morphine. The tests performed show only no evidence of effect, not evidence of no effect |
| 5 | Bruera et al (2004) | Methadone vs morphine (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed |
2 test, Pearsons rho, Wilcoxon rank-sum test, Fisher's exact test | Tests appear appropriate | |
| 6 | Coluzzi et al, 2001 | OTFC vs IR morphine for breakthrough pain (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed | Three-way-ANOVA | Tests appear appropriate | |
| 7 | Cundiff et al (1989) | CR morphine vs IR morphine (E) | Pain intensity | No sample-size calculations mentioned | Two-way ANOVA, parallel line log-ratio assay (some kind of ANOVA) | Old reference (Finney), difficult to distinguish the method from other ANOVA | |
| 8 | Deschamps et al, 1992 | IR release vs CR release morphine (E) | Pain intensity | No sample-size calculations mentioned | Repeated-measures ANOVA, paired t-test | Tests appear appropriate | |
| 9 | Finn et al (1993) | SR morphine tablets compared with IR morphine solution (E) | Pain intensity | No sample-size calculations mentioned | Linear regression, McNemars test, ANOVA | Tests appear appropriate | |
| 10 | Gabrail et al (2004) | ER oxymorphone vs CR oxycodone (E) | BPI (pain intensity and interference) | No sample-size calculations mentioned | Mixed-effects model | The authors ignore a trend because it is stated to be 'not clinically significant'. This is not supported by analyses | Authors conclude that oxymorphone ER and oxycodone CR were considered equivalent if the confidence interval around the treatment difference included zero. This kind of two-sided test can only tell whether the two are different, not whether they are equivalent |
| 11 | Gillette et al (1997) | Oral morphine syrup vs SR morphine capsules (E) | Pain intensity | No sample-size calculations mentioned | Linear regression, Spearman's rank order correlation test | A 'test' for bioequivalence is mentioned, however not justified. Other tests appear appropriate | |
| 12 | Hagen and Babul (1997) | CR oxycodone vs CR hydromorphone (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed l | Three-way-ANOVA, two-way ANOVA, Fisher's exact test,
2 test, binomial test | Tests appear appropriate | |
| 13 | Hanks et al (1987) | CR morphine vs IR morphine solution (E) | Pain intensity | No sample-size calculations mentioned | Mann–Whitney U-test, 'standard crossover-design nonparametric techniques' (two-sample t-test) | Six different outcomes were tested, no adjustments were performed | |
| 14 | Hanks et al (1995) | SR morphine tablet (200 mg) vs two 100 mg tablets (E) | Pain intensity | No sample-size calculations mentioned | ANOVA, Wilcoxon signed-rank test, trapezoidal method for AUC, t-test | The statistical analyses seem appropriate | |
| 15 | Hays et al (1994) | CR hydro-morphone vs IR hydro-morphone (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed | Three-way ANOVA | Tests appear appropriate | Authors conclude that CR hydromorphone is as effective as IR hydromorphone. The tests performed show only no evidence of effect, not evidence of no effect |
| 16 | Heiskanen and Kalso (1997) | CR oxycodone vs CR morphine (E) | Pain intensity | No sample-size calculations mentioned | Mann–Whitney U-test, Wilcoxon signed-rank test, paired t-test,
2 test, regression analysis, one-way and two-way crossover ANOVA | Tests appear appropriate. Not possible to ascertain which tests used at what time | |
| 17 | Hoskin et al, 1989 | CR morphine+IR morphine vs CR morphine+placebo | Pain intensity | No sample-size calculations mentioned | Trapezoidal method for AUC, regression (least squares), t-test | Tests appear appropriate | |
| 18 | Kalso and Vainio (1990) | Morphine vs oxycodone (E) | Pain intensity | No sample-size calculations mentioned | Wilcoxon signed-rank test, rank-sum test, t-test, Spearmans rank correlations, linear regression | Tests appear appropriate | |
| 19 | Kaplan et al (1998) | CR oxycodone vs IR oxycodone (E) | Pain intensity | No sample-size calculations mentioned | ANOVA (two-way, repeated measures), Fisher's exact test, Kruskal–Wallis test | Tests appear appropriate | Authors conclude that CR oxycodone is as effective as IR oxycodone. The tests performed show only no evidence of effect, not evidence of no effect |
| 20 | Klepstad et al (2003) | SR morphine vs IR morphine (E) | Time needed to achieve pain relief | Appropriate pre-hoc calculations appear to have been performed | t-test, Mann–Whitney U-test | Tests appear appropriate | Authors conclude that SRM given daily and IRM given 4-hourly are equally effective. The tests performed show only no evidence of effect, not evidence of no effect |
| 21 | Knudsen et al (1985) | SR morphine tablets vs IR morphine tablets or suspension (E) | Pain intensity | No sample-size calculations mentioned | Wilcoxon paired rank-sum test | Tests appear appropriate | Difficult to understand why the significant finding is not clinically meaningful |
| 22 | Lauretti et al (2003) | SR morphine vs SR oxycodone (E) | Consumption of rescue medication | No sample-size calculations mentioned | Mann–Whtiney U-test, Wilcoxon signed-rank test,
2 test | Adverse effects analysed with
2 test, this is not appropriate on small samples | |
| 23 | Melzack et al (1979) | Brompton mixture vs morphine (E) | Pain intensity | Stated that a subject group of 20 is substantial in a crossover design. No sample-size calculations mentioned |
t-test,
2 test | Tests appear appropriate | No reason stated for choosing P-value of 0.01, however, several outcomes were tested, therefore appropriate to use a lower level |
| 24 | Mignault et al (1995) | SR morphine (MSC) 8-hourly vs 12-hourly administration (E) | Pain intensity | Some posterior power calculations performed | Pairwise t-test, McNemars test | Tests appear appropriate | |
| 25 | Moriarty et al (1999) | CR hydromorphone vs CR morphine (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed | Koch nonparametric method for crossover studies, binomial test | Tests appear appropriate | Authors conclude that hydromorphone and morphine are equally effective. The tests performed show only no evidence of effect, not evidence of no effect |
| 26 | Mucci-LoRusso et al (1998) | CR oxycodone vs CR morphine (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed | Two-way ANOVA, Kaplan–Meier and Logrank-test, Fishers exact test, linear regression | Tests appear appropriate | Authors mention Kaplan–Meier estimate and log-rank test under 'statistical analysis'. Results of these analyses unclear |
| 27 | O'Brien et al (1997) | MXL morphine dosed once daily vs MST continuous dosed twice daily (E) | Use of escape medication | No sample-size calculations mentioned | Double triangular sequential test, Koch method for crossover studies, McNemars test,
2 test, binomial test | Stated that the study should have stopped after 33 patients. However, continued until 69 patients. This may be against protocol | |
| 28 | Parris et al (1998) | CR oxycodone vs IR oxycodone (E) | Pain intensity | Appropriate pre-hoc calculations appear to have been performed | Fisher's exact test, two-way ANOVA, two-way ANCOVA | Tests appear appropriate | |
| 29 | Portenoy et al (1989) | SR morphine tablet (100 mg) vs three 30 mg tablets (E) | Pain intensity | Some posterior power calculations performed |
t-test,
2 test, repeated-measures ANOVA, Fisher's exact test, Wilcoxon rank sum test | Tests appear appropriate | |
| 30 | Stambaugh et al (2001) | CR oxycodone every 12 h vs IR oxycodone given qid (E) | Pain intensity | A comment on sample size was presented, however, it remains unclear whether any calculations were performed | ANOVA, signed rank test | Tests appear appropriate | |
| 31 | Thirlwell et al (1989) | Oral morphine solution vs CR morphine tablets (E) | Pain intensity | No sample-size calculations mentioned | Trapezoidal method for AUC, t-test, repeated-measures ANOVA, Wilcoxon signed-rank test, linear regression | Tests appear appropriate. The adjustment of the significance level due to large number of comparisons is appropriate | |
| 32 | Walsh (1985) | Oral aqueous solution of morphine compared to SR morphine tablets (E) | Pain intensity | No sample-size calculations mentioned | Paired and unpaired t-test | Results of analyses not presented | |
| 33 | Walsh et al (1992) | SR morphine dosed every 12 h vs IR morphine dosed 4 hourly (E) | Pain intensity | Some posterior power calculations performed | ANOVA,
2 test, McNemars test | Tests appear appropriate | Authors conclude that SRMS is as effective as IRMS. The tests performed show only no evidence of effect, not evidence of no effect |
| 34 | Wilder-Smith et al (1994) | Tramadol vs morphine (E) | Pain intensity | No sample-size calculations mentioned | Wilcoxon signed-rank test | The adjustment of the significance level due to large number of comparisons is appropriate, as are the statistical analyses |
(E)=equivalency study; (ER)=extended release; (CR)=controlled release; (IR)=immediate release; (SR)=slow –release; BPI=Brief Pain Inventory; ANOVA=analysis of variance; AUC=area under the curve; OTFC=oral transmucosal fentanyl citrate; CRM=controlled-release morphine.
