PENNSYLVANIA STATE UNIV on March 4, 2016 At the conclusion of the final treatment and after completing the anxiety inventories and the VAS, each subject completed the exit questionnaire. FINDINGS The first research question addressed the pain experienced by the subjects before and after each of the four treatments. This question was tested using an ANOVA, which yielded a significant finding of F = 12.890, p < .0001. A Scheffé post hoc test (∝ = .05) indicated that each pretreatment pain score, as measured by the VAS, was significantly higher than the posttreatment pain score (p < .0001). In addition, the pain score reported before the first treatment was significantly higher than the pain score reported before the second treatment (p < .0001), the pain score before the second treatment was significantly higher than the pain score reported before the third treatment (p < .001), and the pain score reported before the third treatment was significantly higher than the pain score reported before the fourth treatment (p < .001). The second research question addressed the pain experienced by the subjects at the completion of each treatment and 2 days following the treatment. This question was tested using an ANOVA, which yielded a significant finding of F = 11.69, p < .01. AScheffé post hoc test indicated that each posttreatment pain score was significantly lower than the score obtained 2 days later, with the exception of the last treatment, where there was no difference between the posttreatment pain score and the pain reported 2 days later. The decreasing nature of the pain reported by the subjects throughout the entire study, as demonstrated by the mean pain (VAS) scores at each observation, is evident in Figure 1, which incorporates the reported pain scores addressed in Research Questions 1 and 2. The third research question addressed the anxiety experienced by the participants before and after each of the four treatments. This question was tested using ANOVA. The ANOVA performed using the scores from the trait anxiety inventory indicated no significant differences between measures throughout the study, F = 1.715, p = .1036. The ANOVAperformed using the scores from the state anxiety inventory yielded a significant finding of F = 5.397, p < .0001. A Scheffé post hoc test (∝ = .05) indicated that the state anxiety scores were significantly higher before each treatment than they were following the Brady et al. / SHIATSU AND LOWER BACK PAIN 63 Downloaded from jhn.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 treatment (p < .001). The nature of the state anxiety of the subjects across the four treatments, as measured by the means of the state anxiety inventory, is evident in Figure 2. The fourth research question, which asked if various attribute variables affected the findings regarding pain and anxiety, was addressed using each demographic variable as a blocking variable and testing the state, trait, and VAS results using ANOVA. The group was divid64 JOURNAL OF HOLISTIC NURSING / March 2001 Figure 1: Means of Visual Analogue Scale scores for total sample Figure 2: Means of state anxiety scores for total sample Downloaded from jhn.sagepub.com at PENNSYLVANIA STATE UNIV on March 4, 2016 ed equally by age using a median split (median = 39.7); subjects were compared above and below 40 years of age. Those individuals who were taking medications for pain prior to the study (n = 29) were compared to those who were not taking medications (n = 37). Hobbies were divided into categories of high-, medium-, and low-impact activity. There were 16 subjects who had high-impact hobbies (running, hiking, and weight training), 28 who had medium-impact hobbies (gardening, golf, walking, and swimming), and 18 with low-impact hobbies such as sewing, reading, crafts, and computers. Employment status was divided into two groups, professional (n = 36) and other (n = 26). Four years of lower back pain was the median length of history of pain for this sample, so the subjects were divided equally using the median to determine if history with pain influenced the pain or anxiety results. Table 1 indicates that there was no difference between the groups in state anxiety scores, trait anxiety scores, or pain (VAS) scores when the subjects were divided according to the various demographic variables. There were 13 participants who had experienced shiatsu prior to the study. When these 13 participants were compared to the other 56 on state anxiety, trait anxiety, and VAS scores, a Mann-Whitney U was used due to the low number in the group who had experienced shiatsu. The results of the Mann-Whitney U indicated that the two groups did not differ on state anxiety scores (z = –1.694, p = .0904), trait anxiety scores (z = –1.630, p = .1032), or pain (VA