In total 103 patients participated in the study of whom 52 were randomly allocated to the massage group and 51 to the control group. Two patients allocated to the massage group withdrew at baseline, one because they received morphine just prior to starting the study and the other because they were unable to understand the assessment scales, thus the analysis reports results for the 101 patients who completed the baseline assessments. The patients were aged between 21 and 81 years, with mean age 53.4 years (SD 13.3 years) and 58.4% were female. The mean time that the patients had experienced chronic pain was 10.4 years (SD 8.94 years). 87% of the patients had a diagnosis which included back pain and 58% had more than one pain site. At baseline, 60.4% had moderate and 39.6% had severe pain as recorded on the verbal rating scale.
Figure shows the number of patients still taking part in the study at each assessment point. After the first hour the drop out from both groups, but particularly from the control group was rapid. By the 2 hour post-treatment assessment only 23% of patients in the control group remained in the study and by the 3 hour assessment only 10% of the control group were available for assessment. In comparison for the massage group 72% of patients remained in the study at 2 hours post-treatment, 54% at 3 hours and 36% at 4 hours post treatment. Because of the very high withdrawal rate at 2, 3 and 4 hours after treatment, the results from these time periods are impossible to interpret and are therefore not presented.
Participant Flow at each stage of study.
The assessment of the patients at all time points was by an observer who was blind to the treatment that the patient received (DC). Occasionally the patient would break the blinding by comments that they made to the observer. For 89 of the 100 patients in which a record of blinding was made, the observer remained blind throughout the assessment, so blinding was maintained in 89% of patients.
The demographic characteristics and baseline scores for the massage and control groups are compared in Table . Comparisons between massage and control groups were tested using chi-squared or two sample t-tests as appropriate. There are no significant differences in the age, gender mix or duration of pain for the two treatment groups.
Comparison of the characteristics of the massage and control groups at baseline (pre-treatment)
The pain verbal rating scale scores are "moderate" or "severe" for all patients, with 40% of the massage group and 39% of the control group recording severe pain at baseline. The similarity of the distribution of pain scores between the two groups is shown in the boxplot in Figure .
Baseline (pre-treatment) pain VAS scores for the massage and control groups.
Comparisons post-treatment of pain VAS scores are shown in Figure and Table . Figure illustrates that the 95% confidence intervals for mean pain VAS for the massage and control groups overlap at baseline, the means are not significantly different. However, the confidence intervals are clearly distinct both immediately post treatment and at 1 hour post treatment, consistent with the massage group having significantly lower mean pain VAS. For the control group mean pain VAS for those who continue beyond 1 hour (the minority) is 36.4, but mean pain VAS for those in the control group who withdraw at 1 hour (the majority) is 74.7, a significantly higher mean (p < 0.001).
Mean and 95% confidence interval for the mean pain VAS scores for the massage group (- - - - -) and the control group (―) at three time points.
Comparison of massage group and control group for both pain VAS scores and changes in pain VAS score from baseline, at post treatment assessments*.
Table compares the baseline score to that immediately post treatment, the massage group experience a statistically significant (p < 0.001 using paired t-test) mean reduction in pain of 16.7 mm (SD 21.2). For the control group the mean change in pain score is -0.04 mm (SD 16.0) which is not significant (p = 0.985 using paired t-test). For those in the massage group there is a significantly greater reduction in pain score both immediately post treatment and at 1 hour post treatment than in the control group. The 95% confidence interval for the difference in mean pain reduction at one hour post treatment between the massage and control groups is 5.47 mm to 24.70 mm, that is the massage group can expect on average to benefit by between 5.47 mm and 24.70 mm greater reduction than the control group up to one hour post treatment.
Figure illustrates the reduction in pain VAS score for the massage group and the negligible change in pain VAS score for the majority of members of the control group. For both groups the median is central in the box of the boxplot, thus supporting the assumption that the data is normally distributed and that a t-test can be used to compare these changes.
Change in pain VAS score between baseline and immediately post treatment for both the massage group and the control group. The change is calculated as baseline – post-treatment, so positive values indicate pain has been reduced by treatment.
The results from the four point verbal rating pain, the four point pain distress scale and the McGill Pain Questionnaire produce the same findings as the pain VAS scale. That is, statistically significant differences between massage and control groups immediately and one hour post treatment.
An example of these similarities is shown in comparisons of McGill Pain Questionnaire scores from baseline to immediately post treatment. The massage group experience a statistically significant (p < 0.001 using paired t-test) mean reduction in McGill pain score of 11.5 (SD 9.93). For the control group the mean change in McGill pain score is 1.22 (SD 7.16) which is not statistically significant (p = 0.237 using paired t-test).
Table showed, using a chi-squared test, a significant association between treatment group and whether the patient gets 50% pain relief, with only one person in the control group but 18/31 and 18/32 patients in the massage group reporting 50% pain relief immediately post and 1 hour post treatment.
Comparison of massage group and control group for 50% pain relief at post treatment assessment.
Table illustrates results from the Spielberger short-form anxiety scale. It shows that comparing the baseline score to that immediately post treatment, the massage group experience a statistically significant (p < 0.001 using paired t-test) mean reduction in STAI score of 3.57 (SD 3.02). For the control group the mean change in STAI score is 0.00 (SD 3.39) which is not statistically significant (p = 1.00 using paired t-test). For those in the massage group there is a statistically significantly greater reduction in anxiety score both immediately post treatment and at 1 hour post treatment than in the control group. The 95% confidence interval for the difference in mean anxiety score reduction at one hour post treatment between the massage and control groups is 1.67 to 4.36 that is the massage group can expect on average to benefit by between 1.67 and 4.36 greater reduction in anxiety score than the control group up to one hour post treatment.
Comparison of massage group and control group for both Spielberger STAI scores and changes in STAI score from baseline, at post treatment assessments.
At the final assessment for the patient, either at 4 hours post treatment or at the point that they requested an analgesic and thus withdrew from the study the patient was asked to assess whether the treatment they had received was poor, fair, good, very good or excellent. In the massage group, 34 rated the treatment as good, very good or excellent, 12 as fair and 4 as poor. In the control group, 4 rated it as good, very good or excellent, 14 as fair and 33 as poor. A chi-squared test for the patients' assessment of the treatment demonstrates that those receiving the massage rate this treatment statistically significantly better than the control group (X2 = 46.6, df = 2, p < 0.001).
At the final assessment the patient was asked whether they would like to have the treatment again. A significantly higher proportion of those in the massage group (X2 = 24.4, df = 1, p < 0.001) said they would like the treatment again. Of the 49 patients in the massage group who were asked this question all 49 (100%) said yes they would like the treatment again. Of the 42 patients in the control group asked this question, 25 (60%) said yes they would like the treatment again.