Changes in Daily Pain Intensity
The means and standard deviations for the pretreatment, posttreatment, and 3-month follow-up daily pain composites are presented in . A significant, F(2, 19) = 4.08, p < .05, Time × Treatment Condition interaction indicated significant differences between the two treatment conditions in change in daily pain over time. Subsequent ANOVAs for each treatment condition separately showed a statistically significant change in daily pain over the three assessment periods for the HYP condition, F(2, 13) = 9.96, p < .001, but not the PMR condition, F(2, 5) = 0.99, p = ns. Univariate analyses showed a statistically significant pre- to posttreatment decrease in daily pain for the HYP participants, t(14) = 4.63, p < .001, but not for the PMR participants, t(6) = 0.11, p = ns. Moreover, although there was a slight increase in daily pain for the HYP participants from posttreatment to follow-up, this increase was not statistically significant, t(14) = 1.07, p = ns, and the decrease in daily pain-intensity scores between pretreatment and 3-month follow-up remained statistically significant, t(14) = 3.02, p < .01, among the HYP participants. However, among the PMR participants, neither the slight decrease in daily pain from posttreatment to follow-up, nor the difference between pretreatment and follow-up daily pain were statistically significant, ts(6) = 1.47 and 1.31, both ps = ns, respectively.
Means and SDs for the Daily Pain Intensity Composite and Pain Interference Scores at Pretreatment, Posttreatment, and 3-Month Follow-Up
|Treatment Condition||Pretreatment||Posttreatment||3-Month Follow-up||F (df) for|
|F(df, for Time × Condition|
|Daily pain-intensity composite|
|Hypnosis||4.55a||1.35||3.17b||1.75||3.48b||2.04||9.97 (2, 13)***||4.08 (2, 19)*|
|Relaxation||4.08a||1.38||4.13a||1.69||3.35a||1.92||0.99 (2, 5)|
|Pain interference (modified BDI score)|
|Hypnosis||4.66a||1.87||3.16b||2.41||3.78b||2.13||7.62 (2, 12) ***||3.29 (2, 17)†|
Open in a separate window
Note. Means with different subscripts are significantly different (p < .05) from one another.
†p < .10.
**p < .01.
***p < .001.
In terms of the rates of clinically meaningful change in daily pain, 7 (47%) of the HYP participants and 1 (14%) of the PMR participants reported a meaningful decrease in daily pain from pre- to posttreatment. These numbers were 7 (47%) and 2 (29%) at the 3-month follow-up for the HYP and PMR participants, respectively.
Changes in Pain Interference
The means and standard deviations for the pretreatment, posttreatment, and 3-month follow-up pain-interference scores are also listed in . The ANOVAs indicated a nonsignificant trend, F(2, 19) = 3.26, p < .10, for the Time × Treatment Condition interaction. Subsequent ANOVAs for each condition separately indicated a significant change in pain interference over time for the HYP condition, F(2, 12) = 7.62, p < .001, but not the PMR condition, F(2,4) = 1.47, p = ns. Univariate analyses showed a statistically significant pre- to posttreatment decrease in pain interference for the HYP participants, t(13) = 4.06, p < .001, but not for the PMR participants, t(5) = 0.48, p = ns. As with daily pain intensity, although there was a slight increase in pain interference for the HYP participants from posttreatment to follow-up, this increase was not statistically significant, t(13) = 1.25, p = ns, and the difference between pretreatment and follow-up pain interference remained statistically significant, t(13) = 2.19, p < .05. The slight decrease in pain interference from posttreatment to follow-up reported by the PMR participants was not statistically significant, t(5) = 1.78, p = ns, nor was the difference between pretreatment and follow-up pain interference, t(5) = 0.28, p = ns.
There are a number of findings from this study that warrant discussion. First, we found that individuals with MS and chronic pain who received a self-hypnosis training intervention reported significantly more benefits from treatment than individuals assigned to a progressive muscle relaxation condition, despite similar treatment outcome expectancies of the participants in the two conditions. Two other important findings concern the prediction of treatment outcome and the use and reported effects of continued self-hypnosis practice after treatment.
Perhaps the largest challenge in designing methodologically sound hypnosis clinical trials is the selection of the control condition (Jensen & Patterson, 2005). A number of control conditions have been used in published hypnosis trials, such as wait-list controls, standard care, and other (active or effective) treatments, among others. Because each of these controls for different possible confounds, any study that uses one or more of these control conditions contributes to our understanding of the specific and nonspecific effects of hypnotic interventions.
Although this study was quasi-experimental because it did not include randomization of all participants, we were able to compare self-hypnosis training to a PMR intervention. This intervention was designed to meet the need to control for treatment-outcome expectancies. Like the hypnosis treatment, it was based on an intervention that has demonstrated efficacy for treating chronic pain, could be described in a way that elicited positive outcome expectancies, could be labeled similarly to the hypnosis treatment (i.e., as an intervention that includes “both hypnosis and relaxation components”) and could be provided in a way that was also very similar to the hypnotic intervention (e.g., face-to-face in 10 sessions, with an accompanying audio recording, etc.). However, the PMR condition in this study differed from the HYP condition in several critical ways, the most important of which was the fact that the PMR condition consisted of only one (but constantly repeated) direct suggestion: to experience relaxation in specific areas of the body. The hypnotic intervention, on the other hand, included a hypnotic induction followed by a much larger number and variety of suggestions, including, in the first two sessions: (a) a suggestion to experience being in a “special place,” (b) a classic hypnotic suggestion to encourage confidence in responsivity, (c) five different analgesia suggestions, (d) posthypnotic suggestions that the benefits obtained with treatment will last beyond the session and become permanent, as well as (e) any additional suggestion that the participant might want to hear during the sessions (to facilitate greater involvement in the sessions and tailoring of treatment). The five analgesia suggestions provided in the first two sessions were reduced to two suggestions (a suggestion for decreased pain unpleasantness plus whatever other analgesia suggestion the participants appeared to enjoy the most or get the most out of), but all of the other suggestions continued for the remaining eight sessions.
Thus, the HYP and PMR treatment conditions shared many key nonspecific components, including their effects on outcome expectancy but differed with respect to the number and variety of suggestions offered. Moreover, both interventions had similar effects on outcome expectancies. Given the fact that the hypnotic-analgesia protocol was more effective than the PMR comparison condition, the findings suggest (but only suggest; see discussion of limitations of quasi-experimental designs, below) that the hypnotic suggestions included in the HYP treatment had an effect on these outcome variables over and above the effects of therapist attention, time, or outcome expectancy.
We examined two predictors of treatment response in this study: hypnotizability and treatment-outcome expectancy. Of these two, only treatment-outcome expectancy was associated (moderately to strongly) with outcome. The lack of a significant association between hypnotizability and treatment outcome is inconsistent with some previous findings in clinical settings (e.g., Andreychuk & Skriver, 1975; Friedman & Taub, 1984; Gay, Philipport, & Luminet, 2002) but is consistent with a previous study by our group using a similar treatment protocol (Jensen et al., 2005). The inconsistencies across studies concerning the relative importance of hypnotizability as a predictor may be related to differences between studies in the way that hypnotizability is assessed, differences in the treatment protocols used, differences in the samples or types of pain studied, or some combination of these.
Even when significant associations between hypnotizability and treatment outcome are found, however, they are not always strong for all outcome measures (Friedman & Taub, 1984; Gay et al., 2002). The skills needed to respond to hypnotic suggestions for pain management, even in the best of circumstances, may not always be strongly related to the skills necessary to respond to the hypnotic suggestions contained in common hypnotizability tests, such as suggestions for arm levitation, amnesia, or visual hallucinations. Hypnotic responding is not necessarily a single unified trait and may be composed of multiple abilities (cf. Pekala & Kumar, 2007), some of which may be associated with response to analgesia suggestions and others of which may not. In any case, as a group, these findings suggest that it is probably not useful to screen individuals from hypnotic treatment for chronic pain management based on their response to hypnotizability tests alone. Such screening may, in fact, exclude some patients from a treatment they could benefit from.
On the other hand, treatment-outcome expectancy did show a moderate to strong association with treatment outcome in this study. Although the present findings do not support a conclusion that the effects of self-hypnosis training are entirely due to expectancy effects (otherwise, we would have seen a similar treatment effect for the two conditions), the findings are consistent with the hypothesis that patient expectancies may play a role in both immediate and short-term (at least up to 3 months) outcomes in response to hypnotic analgesia treatment for chronic pain (Kirsch, 1985); although the importance of outcome expectancies in hypnotic responding may be much less than is commonly thought (Benham, Woody, Wilson, & Nash, 2006). Practically, the findings suggest the possibility that clinicians might be able to enhance treatment outcome to some extent by presenting treatment in a way that realistically describes treatment and its possible effects and also facilitates patient expectancies and hope for positive outcomes. This possibility certainly warrants further investigation. Research that identifies ways to enhance outcome expectancies in the clinical setting, and then determines the impact of this enhancement on clinical outcome, could be particularly useful to help (a) provide additional tests of the relative importance of expectancy in determining response to hypnotic treatment and (b) possibly enhance the efficacy of hypnotic analgesia treatment.
Three limitations of the current study should be kept in mind when interpreting the results: (a) the quasi-experimental design; (b) the low sample size; and (c) the “active” nature of the comparison (PMR) condition. Although it was important to include 8 of the participants in the analyses of the HYP condition who had been given HYP from the start in order to increase the power of the analyses, the inclusion of these participants in the analyses limits our ability to draw causal conclusions about the effects of HYP versus PMR from the study. Future research, ideally with larger sample sizes, will be needed to determine the extent to which the findings replicate to other samples. In addition, the low sample size limits the ability of the study to detect effects that might exist in the population but did not emerge in the sample. For example, although both interventions resulted in reduced pain, only the reduction observed in the HYP condition was statistically significant. It is possible that the pain reduction reported by participants in the PMR condition might have been found to be statistically significant had there been a larger number of participants in the study who received PMR. Similarly, some of the differences observed between the two treatment conditions concerning the effects of practice on pain (for example, that the HYP recordings reportedly resulted in more hours of pain relief than the PMR recordings did) might have been statistically significant had we had more resources to recruit additional participants for the study. Future researchers should strive to maximize the numbers of participants in hypnosis clinical trials to be better able to detect true effects or to be more confident that such effects do not exist when a lack of significant difference is found.
The strengths of the PMR comparison condition we used in this study have already been discussed. But all comparison or control conditions used in hypnosis studies have both strengths and weaknesses. A primary weakness of the PMR condition used in this study, already alluded to, is that it is an active condition that, in fact, may benefit individuals via similar mechanisms as hypnosis. PMR has been found to be effective for pain management in other studies (e.g., Baird & Sands, 2004; Crockett, Foreman, Alden, & Blasberg, 1986) and was associated with a reduction in pain (at least from pre- to postsession) in the current study. Because PMR is an active (and potentially effective) treatment, the differences noted between HYP and PMR in this study may underestimate the actual effects of the HYP intervention if compared to an inactive control or no treatment. It is difficult to isolate the unique components of the HYP intervention from those we might employ in a comparison condition. Thus, although the comparison condition was useful for testing and confirming an effect of the hypnosis treatment over and above the effects of time, therapist attention, and patient expectancy, because it is an active treatment we may interpret the results in ways that understate the effectiveness of the hypnosis treatment. For this reason, the PMR condition is not useful for determining the effects of hypnosis relative to no treatment or “nonhypnotic” care. Estimating these treatment effects would have required a third condition, such as a wait-list control. Future researchers would be wise to include such a condition whenever possible; although we understand that the resources available for conducting a clinical trial are often limited, and that the requirements for statistical power may require a limitation in the number of treatment conditions offered in any one study.
We have previously argued that no single hypnosis clinical trial can be definitive, and there is no such thing as a perfect control condition for hypnosis studies (Jensen & Patterson, 2005). Rather, in order for our understanding of the effects of hypnosis on pain and other conditions to advance, the field requires multiple clinical trials and studies that compare hypnotic interventions to a variety of control conditions and interventions. Ultimately, such a series of studies will produce a body of evidence that can help to clarify the efficacy and impact of hypnosis on pain and other symptoms. For the present, however, the results we report here encourage further examination of the clinical utility of hypnotic methods for chronic pain management.