Controlled trial of Internet-based treatment with telephone support for chronic back pain

1 Introduction

Pain management is often expensive given the extensive need of rehabilitation (^{Swedish Council on Technology Assessment in Health Care, 2000}). Being one of the most common causes of disability and sick leave, there is a need to develop alternative cost-effective ways to administer multidisciplinary treatments (^{Turk and Okifuji, 2002}). It is fairly well established that psychological input is considered necessary in rehabilitation, and in particular, cognitive-behavioral approaches (^{Morley et al., 1999}). However, for many people with back pain, cognitive-behavioral treatment is not available (^{Turk and Okifuji, 2002}). An alternative approach is to help patients treat themselves by means such as self-help books supplemented with feedback from a therapist, possibly in the form of telephone calls. For several clinical problems, such as headache (^{Haddock et al., 1997}), this has turned out to be a successful approach. Apart from headache, self-management of chronic pain has not been subjected to much research. In the few studies that are available, findings have varied, with some studies showing a substantial dropout rate (^{Spence and Sharpe, 1993}).

Recently, a new form of self-help treatment has been developed and tested using the Internet to present self-help material and individualized feedback in the form of e-mail correspondence as means of interaction with the user (^{Ritterband et al., 2003a}). So far, the effectiveness of the Internet-based self-help treatment has been investigated in studies on recurrent headache (^{Andersson et al., 2003b; Ström et al., 2000}), tinnitus (^{Andersson et al., 2002}), panic disorder (^{Carlbring et al., 2001; Richards and Alvarenga, 2002}), post-traumatic stress (^{Lange et al., 2001}), cancer (^{Winzelberg et al., 2003}), stress management (^{Zetterqvist et al., 2003}), weight loss (^{Tate et al., 2001}), pediatric encopresis (^{Ritterband et al., 2003b}), and eating disorders (^{Winzelberg et al., 2000}). The results from these studies are promising and indicate that the Internet is a cost-effective medium for providing services and that the effects are close to the effects of face-to-face treatment. In a recent study, positive effects were found when back pain patients participated in a closed, moderated, e-mail discussion group (^{Lorig et al., 2002}).

The aim of this study was to investigate whether previous pain management techniques, derived from the cognitive behavioral literature, could be delivered using Internet technology. The study was set up as a randomized controlled trial and all components were delivered via the Internet, including recruitment, assessment, treatment delivery, and follow-up.

2 Methods

2.1 Participants

Participants were recruited by means of newspaper articles in national and regional papers as well as through a Webpage for health on the Internet. Webpages providing an outline of the study as well as an application and screening form, were created as a way to recruit subjects. Subjects were recruited between the 3rd and 21st of January 2001.

A total of 100 people showed an interest in the study, and 67 fulfilled the inclusion criteria. Inclusion and exclusion criteria are presented in Table 1. A physician was consulted to exclude persons in accordance with the criteria “suffer from pain that can increase as a consequence of activity (e.g. spinal stenosis).” The assessment was made on the basis of the participant's self-report. An inclusion criteria was hence to have seen a physician but no verification of illness was obtained. However, participants were requested to provide their diagnosis and medical history. A physician reviewed this information. Individuals who mentioned that they suffered from depression received an e-mail informing them that the treatment was exclusively for pain patients. In order to exclude those in great need of depression treatment, each person who mentioned depression as a symptom was also asked to describe their last depressive episode. However, no one was excluded due to depression.

Of the 67 eligible participants who met inclusion criteria, 56 expressed interest to participate. Of these 56 participants, 5 (8%) dropped out at various stages without providing post-treatment data, leaving a final sample of 51 participants (35 female, 21 male), 22 in the treatment condition and 29 in the control condition. Student's t-test showed that the dropouts did not differ significantly from those completed in pain duration or age. Table 2 presents descriptive data on demographic and other relevant characteristics for the participants who completed the program and for the dropouts. There were no significant between-group differences. Mean age was 44.6 years (SD=10.4). Participants suffered from pain for a mean length of 10.1 years (SD=9.2), and 32.1% were on sick leave.

2.2 Design

Following recruitment and informed consent (which was obtained electronically) participants were assessed and randomly assigned to one of two conditions: (a) an Internet-based pain management program or (b) waiting-list control. Data were collected at pre- and post-treatment and were analyzed in a mixed design with one between group factor and one within group factor. Follow-up data were collected at 3 months post intervention. Data were obtained from 47 of the 51 participants (92%). There was 1 dropout in the treatment group and 3 in the former control group. The control group received the treatment on a more open basis with one password for the whole program. All took part in the program except the dropouts.

3 Material

3.1 Inclusion form

The inclusion form contained 22 questions about the participants' pain, including what affects their pain, other treatments utilized and other possible diagnoses. The inclusion form also included the Hospital Anxiety and Depression Scale (HADS) (^{Zigmond and Snaith, 1983}), and questions regarding age, gender, telephone number and socioeconomic factors, such as workplace and sick leave. The inclusion form was partly derived from Philips and Rachman's manual (^{Philips and Rachman, 1996}) and from ^{Ström et al. (2000)}.

3.2 Key outcome variables

The rationale for choosing outcome measures was based on previous studies on CBT for chronic pain and the objective to be able to compare outcomes. The primary outcome measure was the Coping Strategies Questionnaire (CSQ; ^{Rosenstiel and Keefe, 1983}). However, the Multidimensional Pain Inventory (MPI; ^{Kerns et al., 1985}), the Pain Impairment Rating Scale (PAIRS; ^{Riley et al., 1988}), and the HADS (^{Zigmond and Snaith, 1983}) were also included as secondary outcome measures.

3.2.1 The Coping Strategies Questionnaire

CSQ, is a 50-item questionnaire (^{Rosenstiel and Keefe, 1983}) assessing cognitive and behavioral coping strategies. It covers eight different coping strategies for pain—diverting attention, reinterpreting pain sensation, coping self-statements, ignoring sensations, praying and hoping, catastrophizing, increased behavioral activities and pain behavior. The subscale pain behavior has low inter-item reliability (^{Jensen and Linton, 1993}) which is why it is often excluded. ^{Jensen and Linton (1993)} recommended the exclusion of the subscale from the Swedish version. At the end of the questionnaire, patients are asked to make two ratings of the overall effectiveness of coping strategies they use. Patients rate (1) how much control they feel they have over pain, and (2) how much they are able to decrease pain. The Swedish version has a high inter-item reliability (α=0.7–0.8), and test–retest reliability ranging from0.4–0.9 (^{Jensen and Linton, 1993}).

3.2.2 Multidimensional Pain Inventory

MPI, assesses psychosocial and behavioral consequences of pain (^{Kerns et al., 1985}), and has been translated into Swedish (MPI-S; ^{Bergström et al., 1998}). The MPI-S is a 34-item questionnaire divided in two sections. The questionnaire consists of 8 scales. These are pain severity, interference, life control, affective distress, support, punishing responses, solicitous responses and distracting responses. The third part has been excluded in the Swedish version. For part one the internal consistency of the MPI-S is α=0.80, for part two α=0.76–0.86 and for part three is α=0.67–0.81 (^{Bergström et al., 1998}). The test–retest coefficients for the MPI-S are r=0.73–0.89. The discriminant validity for part one and two, after excluding four statements, indicates that MPI-S have the same scale structure as the MPI. For section three the scale structure could not be verified why this part was excluded from the Swedish version.

3.2.3 The Pain and Impairment Relationship Scale

PAIRS, consists of 15 personal statements that reflect thoughts, attitudes and opinions about pain (^{Riley et al., 1988}). The questionnaire assesses beliefs and attitudes that patients have regarding pain and ability to function despite discomfort. PAIRS has an internal consistency of α=0.82 (^{Riley et al., 1988}). Test–retest reliability is r=0.66.

3.2.4 Hospital Anxiety and Depression Scale

HADS, is a 14-item questionnaire, and is designed to measure anxiety and depression in non-psychiatric patients treated at hospital clinics (^{Zigmond and Snaith, 1983}). Items referring to physical symptoms (e.g. dizziness or headaches) have been eliminated, since such reactions may also be attributed to the medical disease and treatment itself. The internal consistency for the anxiety scale is α=0.80–0.93 and for the depression scale α=0.81–0.90 (^{Bjelland et al., 2002}). Test–retest reliability after two weeks is r>0.80.

3.3 Pain diary

The participants made daily assessments of their pain intensity for 1 week before and 1 week after going through the intervention. This was done on paper-and-pencil and later put online. Participants assessed their pain intensity on a scale from 0 (no pain at all) to 100 (extremely pain) three times every day (morning, noon and evening). A pain index was computed (the sum of the assessed pain intensity divided with the number of recordings every week). When data were analyzed the highest rated pain intensity for each day was also taken into consideration and used to calculate the mean peak intensity (e.g. the sum of the peak pain intensity ratings divided by the number of peaks for the whole week).

3.4 Treatment credibility

To obtain a measure of treatment credibility, participants were given a 5-item, 10-point scale adapted from ^{Borkovec and Nau (1972)}. The five items dealt with (a) how logical the treatment was perceived to be, (b) certainty that it would be successful, (c) if they would recommend the treatment to a friend with pain, (d) how effective the treatment would be for other problems (not pain), and finally (e) how much improvement they expected to achieve. The credibility scale was completed at the end of the first self-help module, after the participant had received a full description and rationale for the treatment.

3.5 Satisfaction with treatment format

Following treatment and registrations, participants were asked to fill out a questionnaire on aspects of the treatment and use of Internet. They were also asked to evaluate the telephone calls.

3.6 Procedure

Interested individual participants filled out the inclusion form on the Website for this study.

All participants had access to a computer, modem and an Internet connection, and could print out the training instructions. The treatment was for free, but the costs for Internet connectivity was not compensated. All information and treatment were presented in Swedish. The technical requirements were adapted to be as accessible as possible. Hence, participants could access the site from their own home via modem and with a standard browser (e.g. Netscape). A contact e-mail was given for technical support and separate instructions on how to download files were also available on the site. Data was captured actively and required that the participant fill-in forms that were later saved to a database. Data on how often participants logged in was not collected, and participants were encouraged to download text material to read. In fact, it was recommended that subjects print out the information and exercises in most instances (apart from a slide show for applied relaxation).

All contact with the participants was via e-mail and telephone. General information about ethical principles, an introduction to cognitive behavior therapy (CBT), specific information about the study and the application form was accessible to all users of the Internet. The training program, as well as rating scales and other forms, was only accessible with a password provided by e-mail to the participants. In cases when the subject was unable to connect to the Internet program, they were given the option to contact the experimenters. A response was provided within 24h. However, there were few such problems.

Participants were informed about the outline of the study by e-mail and the information on the webpage. They were instructed to monitor their pain intensity on a daily basis for 1 week (Pain Diary) and to fill out the MPI-S, PAIRS and CSQ. The pain diary was submitted once the whole week had been registered. Following this initial monitoring period, participants were randomized into treatment or waiting list conditions. Subjects were randomized using dice, where even numbers meant treatment and odd numbers meant control condition. Participants in the treatment group received a code to enter the intervention and the controls were informed that they were on a waiting list. They were offered the program two months later, after registering an additional period of pain intensity monitoring.

During the treatment phase, participants were instructed to follow the scheduled program, reading the information corresponding to each week, and submitting treatment registrations regularly. If the diary data were not delivered as expected, a reminder was sent one week later. Participants were also encouraged to ask questions or comment on pain, the training program or other relevant issues. They could do this by e-mail or during the weekly telephone call. The therapists were graduate students trained in CBT. They were both given regular supervision by a qualified clinical psychologist who specialized in pain management as well as occasional supervision from a physiotherapist based in the same pain clinic. After completing the whole program, participants monitored their pain intensity for one week and filled out the MPI-S, PAIRS, CSQ and HADS questionnaires again. On the last week, participants in the treatment group also filled out a questionnaire on satisfaction with the treatment format (e.g. use of the Internet and telephone calls). Adherence with the treatment in terms of completion of exercises and other home-work assignments (see Table 3) was checked during each telephone contact and also by means of diary ratings submitted via the Internet.

Finally, a 3-month follow up was conducted in which participants were contacted to fill-in their pain intensity for 1 week and the questionnaires (MPI-S, PAIRS, CSQ and HADS). This was timed to be equal between the two groups.

3.7 Self-help program

The treatment model was mainly derived from a cognitive-behavioral model of chronic pain (^{Vlaeyen and Linton, 2000}). The pain management program used in the study was derived from the cognitive-behavioral literature (e.g., ^{Marhold et al., 2001; Philips and Rachman, 1996; van Tulder et al., 2000}), and included psychological components (e.g. dealing with unhelpful thoughts and beliefs, changing focus) as well as stretching and physical exercises (on an individualized graded activity basis but with structured information). Participants were taught different coping strategies, which was the main component of the program. The aim was to identify more active ways of coping with their pain and to improve their level of functioning. Participants were also offered a program of applied relaxation (^{Öst, 1987}), used in the previous Internet studies (^{Ström et al., 2000}) and in the pain management model developed by ^{Linton (1994)}. Applied relaxation has also been evaluated in earlier studies (e.g. ^{Johansson et al., 1998}). All instructions were adjusted to suit the self-help format and included the option of viewing a slideshow or downloading sound files as an aid to using applied relaxation. Participants were also offered a CD with the instructions for applied relaxation. The program lasted 6 weeks, and participants were prompted to send in their treatment information weekly. The rationale for this data exchange was mainly to keep a regular contact, and thereby promote compliance. The weekly components of the program are summarized in Table 3.

3.8 Telephone contact

The participants spoke on the phone with one of the therapists once a week. The telephone calls were structured and every call consisted of reviewing the previous week's homework, answer questions the participant had and to provide possible reminders and maintain motivation. Below follows a more complete description of the telephone sessions.

During the first telephone call the participants’ goals with the treatment were discussed. The goals were divided into smaller and concrete. The participants' risk situations were also discussed. The relaxation training and possible difficulties were discussed. During the second telephone call the participants asked questions about the relaxation training. Exercise was also discussed. The participants were also reminded of the importance of taking pauses and doing stretch exercises. The third telephone call was about physical exercise and stretching. The relaxation training was also discussed. The participant was able to ask questions about the different coping strategies and the coping diary. During telephone call four and five, all the coping strategies were discussed. The participant was also reminded to use relaxation exercises as often he/she could. During the last telephone call the different goals were evaluated and the participants maintenance program was discussed. A relapse prevention plan was also discussed. Finally an overall review of the treatment was discussed.

3.9 Statistical procedure

Because of the large number of dependent variables, the data was analyzed in two ways. For the inventories comprising subscales multivariate analysis of variance (MANOVA) were conducted on the change scores between pre- and post measurements. Significant between-group effects were followed by univariate ANOVAs. The first set comprised of nine subscales of the CSQ, and the second set included the eight scales of the MPI. The third set included mean scores of the HADS scales (anxiety and depression), PAIRS, and diary ratings of pain intensity and highest pain intensity. As these variables were separate from the others, repeated measures ANOVA were used. For the three-month follow-up repeated measures, ANOVA were used for the whole group, as all had received treatment at that point. Analysis of proportions of participants showing clinical significant improvement was analyzed by means of Chi square, Odds ratio, and numbers needed to treat (NNT) analyses. All analyses were conducted on a per-protocol basis (e.g. not intention-to-treat).

4 Results

Means and standard deviations for all measures are presented in Table 4. There were no significant pretreatment differences (all p>0.10).

4.1 The coping strategies questionnaire (CSQ)

The MANOVA on the CSQ data produced a significant between-group effect, F(9,41)=2.14, p=0.047. Significant univariate effects were obtained for the subscale catastrophizing [F(1,49=7.28, p<0.01], control over pain [F(1,49)=12.20, p<0.001], and ability to decrease pain [F(1,49)=4.61, p<0.05]. No other significant effects were found.

4.2 Multidimensional pain inventory (MPI)

MANOVA did not show a significant effect, F (8,42)=0.6, n.s.

4.3 The pain and impairment relationship scale (PAIRS)

No significant interaction was found for group and time. A significant main effect for time was found [F(1,49)=7.23, p<0.01], that is both groups experienced a reduction in functional impairment and beliefs, and attitudes regarding pain improved regardless of group assignment.

4.4 Hospital anxiety and depression scale (HADS)

No significant main effects or interaction for group and time were found.

4.5 Pain diary

No significant interaction between group and time for average pain intensity or highest rated pain intensity was found. For average pain intensity, there was a significant main effect for time [F(1,48)=6.10, p<0.05], that is the average pain intensity was reduced for both groups over time. A main effect was also found for highest rated pain intensity [F(1,48)=13.01, p<0.001], that is the highest rated pain intensity was reduced for both groups.

4.6 Clinical significant improvement

Given a long-standing interest in describing outcome data in terms of clinical significant change (^{Ogles et al., 2001}), Jacobson's reliable change index was calculated for each individual based on the pre-treatment score, the post-treatment score and the standard error of the difference [Xpost−Xpre]/S diff (^{Jacobson and Truax, 1991}). The calculation requires access to test–retest reliability. The main focus here is on the CSQ variable catastrophizing, for which a significant effect was found and which has adequate reliability (^{Stewart et al., 2001}). The PAIRS was also analyzed in terms of proportion reaching the reliable change criterion. For the catastrophizing scale, 39% (n=10) showed a reliable improvement; 14% (n=4) in the control group, a difference that was statistically significant, χ²(1)=4.4, p<0.05. Expressed as odds ratio, the result was OR=3.9 (95%CI=2.96–4.84), with an NNT of 4 (95%CI=1.65–25.8). For the PAIRS, 27% (n=7) showed a reliable improvement in the treatment group; 10% (n=3) in the control group. This difference did not, however, reach statistical significance χ²(1)=2.53, p=0.11).

4.7 Treatment credibility

The experimental group scores on the treatment credibility scale (M=36.8, SD=5.5) were correlated with their changes on the subscales, in which significant interactions were found (e.g. catastrophizing, control over pain and ability to decrease pain). No significant correlations were found between treatment credibility and change scores on the respective subscale. This indicates that the participants' preceding impression of the treatment program did not have any relation with the treatment's outcome.

4.8 Three-month follow-up

At this stage, all participants had received the treatment and were hence treated as one group. Repeated measures ANOVA showed that there were significant decreases in scores from pre-treatment to follow-up for the following CSQ subscales: praying and hoping [F(1,46)=4.9, p=0.032], catastrophizing [F(1,46)=14.0, p=0.005], control of pain [F(1,46)=37.3, p<0.001], and ability to decrease pain [F(1,47)=21.7, p<0.0001]. For the MPI, a significant improvement over time was found for Life control [F(1,46)=15.7, p<0.001] as well as a decrease of punishing responses [F(1,46)=4.38, p<0.05]. Results on the PAIRS showed a significant reduction [F(1,46)=10.8, p<0.01]. A significant decrease was also found on the HADS depression scale [F(1,47)=11.6, p<0.001].

5 Discussion

In the present study, positive effects were found following a CBT self-help program for chronic pain administered via the Internet. The findings should be seen in the light of the effects of CBT in general for chronic pain (^{Morley et al., 1999}), as the treatment was adapted from a multidisciplinary pain management program, but with a major psychological component. The treatment protocol is different from previous self-help treatments for chronic pain (^{Spence and Sharpe, 1993}) in that the Internet was used to facilitate closer monitoring and feedback during treatment. In addition, telephone support was provided as a way to increase subject motivation and compliance. This has been a significant problem in the previous controlled trials on Internet-based treatments which suffered from substantial dropout rates (^{Ström et al., 2000}). In the present study, the dropout rate was very low given that it is a self-help program, suggesting that telephone consultations may have greatly improved adherence. However, in a recent study on Internet-based self-help treatment for headache, telephone support did not seem to have any influence on dropout rates or the amount of improvement (^{Andersson et al., 2003b}). Obviously, the addition of weekly telephone calls can be seen as a restriction on the cost-effectiveness of Internet treatment. In addition, we cannot exclude the possibility that the telephone calls were at least in some respects responsible for the effects seen.

Based on previous estimates of effect sizes following CBT for chronic pain (^{Morley et al., 1999}), an effect size of at least Cohen's d=0.5 was expected. According to the tables given by ^{Cohen (1988)}, this would require a sample of 95 participants (given a p-value of 0.05, power 80%); 100 participants were recruited for this study. Unfortunately, due to exclusion criteria, this number was not reached. However, other researchers have found slightly higher effect sizes (e.g. ^{Flor et al., 1992}), but the actual power for the obtained sample size in the present study was 45%.

The present study suggests that self-help via the Internet might be an alternative for at least some chronic pain patients. The present adaptation of a behavioral medicine protocol resulted an improvement in the short run (e.g. catastrophizing). For a few measures some improvement was also seen in the control group. This may be explained by participants being in different stages of change (^{Zetterqvist et al., 2003}). The stages of change construct has been explored in the pain management literature (^{Kerns et al., 1997}), and might be relevant in the context of self-help, which to a great extent assumes that the patient is prepared for change.

At 3-month follow-up, the overall pattern suggested that the treatment had beneficial effects. For example, catastrophizing had decreased from pre-treatment to follow-up, with improvements also seen on other measures. Unfortunately, there was no control group data at this stage due to the cross-over design implemented which would have been possible if a non-specific treatment control had been used instead of a waiting-list group.

There are several caveats with the findings here. When recruiting participants for the study, articles appeared in Swedish newspaper journals and on a Website. How representative the participants in this study are in comparison to clinically recruited samples can be debated. It might be that those participating in the intervention were a select group. One might expect these persons to already be committed to making behavior changes and be more motivated. While all participants were self-recruited and not selected from a pain clinic, a broader range of pain patients were likely reached, at least geographically, than would have been if participants had been recruited locally.

A problem with Internet recruitment is that it may be difficult to generalize the results obtained from the sample to the population. All the participants in this study had access to the Internet and had some computer skills. This problem is difficult to control. Although a majority of Swedish citizens have access to the Internet, it is unlikely whether this is the case with chronic pain patients. To ensure that the sample would be representative of the population, inclusion criteria similar to previous studies were used.

While most of the patients that receive clinical treatment are on sick leave, a majority of the participants (67.9%) in this study were not. This could indicate that the participants in this study were less distressed than participants in other trials, or at least more able to work. However, the disability scores of study participants did not differ greatly from those reported in other pain studies. Possibly, the problem sometimes encountered with the competing demands of improvement in terms of getting back to work in relation to the desire to maintain disability benefits was avoided. Indeed, ^{Marhold et al. (2001)} found that improvements resulting from a cognitive-behavioral pain management program only pertained to persons with a history of short term sick leave, but not for those on long-term sick leave. Unfortunately, no data was collected on return to work after the follow-up period.

Outcome was judged solely on subjective self-rated questionnaires administered via the Internet. Although Internet administration of questionnaires has been found to generate reliable and valid data (^{Andersson et al., 2003a; Buchanan and Smith, 1999}), the specific psychometric properties when used with pain patients have not been established. An independent assessment carried out at follow up, or an interview with an informant, might have improved the validity of the assessments. In addition, functional measures would be possible to obtain during such assessment. Open-ended responses were obtained from participants regarding satisfaction with treatment and therapists. Overall, these comments were very positive and did not mirror the more modest effects seen in the statistical analyses.

In conclusion, while there is more to be learned about Internet-based intervention for chronic back pain, this is one of the first studies to investigate the effects of an Internet-based self-help program. The results suggest that Internet-based self-help, based on established psychological treatment methods, holds promise as an effective approach for treating disability in association with pain.

Acknowledgements

This research was supported through a project grant from the Swedish Council for Social Research. We acknowledge the kind assistance of Jaime Navia, Nils Olaisson and Maria Jannert. Assistant Professor Lee M. Ritterband is acknowledged for helpful comments and input on the manuscript.