An unblinded, parallel-arm, pilot randomized controlled trial with equal allocation was conducted. The participants were randomized to the internet-based multimodal pain program or to an active control. The study was approved by the Ethics Review Board in Lund, Sweden (No. 2016/6). The examination and treatments were free of charge for the participants, and no other financial compensation was given. See Multimedia Appendix 1 for the informed consent documentation in Swedish. This study was retrospectively registered at ClinicalTrials.gov (NCT04363762).

Furthermore, this study was part of a large research project with a primary objective of studying changes in the brain in patients with chronic TMD pain after treatment. Therefore, all participants underwent magnetic resonance imaging (MRI) of the brain. The neuroimaging data will be presented elsewhere; however, the feasibility aspect of this additional examination will be addressed here.

Participant inclusion criteria were (1) age between 18 and 75 years; (2) at least one TMD pain diagnosis such myalgia, myofascial pain with referral, headache attributed to TMD, or arthralgia according to the Diagnostic Criteria for Temporomandibular Disorders (DC/TMD) [13]; (3) chronic (≥3 months) TMD pain, experienced once a week or more often, with an intensity of ≥3 (on a scale from 0 to 10); (4) access to a computer with an internet connection and a mobile phone; (5) sufficient computer literacy; and (6) Swedish language fluency. Exclusion criteria were (1) chronic inflammatory systemic diseases; (2) all psychiatric disorders except depression and anxiety due to high comorbidity; (3) occlusal splint therapy in the past 12 months; (4) ongoing extensive dental treatment; and (5) conditions precluding MRI examination.

Participants were recruited from a general dental care clinic (Fäladstorget, Lund) within the National Dental Care in Skåne, Sweden. To identify potential participants, all consecutive adult patients visiting for a regular dental check-up were screened with the 2Q/TMD screening tool [14,15]. If a patient was deemed eligible for the inclusion examination, oral and written information about the study were given, and a new appointment was booked for an inclusion examination. The patient was asked to provide written informed consent, if they were deemed eligible after the inclusion examination. Permuted block randomization with a fixed block size of 10 was used. At the moment of assignment, the local research coordinator blindly picked a piece of paper from an envelope with allocated treatment. Before the study start, 6 sets of opaque envelopes with 10 allocation notes each, 5 with “internet-based multimodal pain program” and 5 with “occlusal splint,” were prepared. All 10 notes in one envelope had to be used before the next envelope was opened. If allocated to internet-based multimodal pain program, working material was sent to the participant by mail, and the treatment was started, with assistance provided by phone. If allocated to treatment with occlusal splint, a new appointment was booked for delivering the splint. Follow-up was performed 3 and 6 months after treatment start with questionnaires that were mailed to the participants. Participants also underwent an MRI of the brain at baseline, after randomization but before treatment start, and 6 months after treatment start at Skåne University Hospital, Malmö, Sweden.

Participant recruitment started in April 2016 and ended in December 2018. Screening was paused during annual summer holiday periods (June-August 2016, 2017, 2018) and during February-April 2017 due to technical issues with the MRI scanner.

All participants underwent a standardized examination according to DC/TMD to control for eligibility and collect baseline data. The DC/TMD includes a clinical examination and questionnaires described below [13]. Demographic data and pretreatment expectancies on allocated treatment were also assessed.

The clinical DC/TMD examination comprises a standardized assessment of pain and headache locations, jaw opening capacity, pain on mandibular movement, pain on palpation, and temporomandibular joint noises [13]. The clinical examinations were carried out by one of 2 trained dentists [16].

The DC/TMD Symptom Questionnaire was used to assess pain symptoms involving the jaw, jaw noise and locking, and headache. Data from this questionnaire were combined with findings from the clinical examination for diagnosis according to DC/TMD [13].

Nonparametric statistics were used due to the characteristics of the pain-related variables. For descriptive statistics, median, 25th percentile, and 75th percentile, as well as the number of observations or proportions, are reported. To explore the potential effects of the internet-based multimodal pain program in this early-stage trial, per protocol between-group comparisons of change in characteristic pain intensity level, pain-related disability, jaw functional limitation score, degree of depression, anxiety, catastrophizing, and stress at the 3- and 6-month follow-up were performed with the Wilcoxon rank-sum test. Friedman analysis of variance was used for within-group comparisons. Significant results from the Friedman analysis were followed by posthoc Wilcoxon matched-pair signed-rank testing. In addition, Spearman rank-order was used to test possible association between positive and negative treatment expectancy and changes in outcome measures at the 6-month follow-up within respective treatment groups. Dropout analysis was performed to test the difference between completers and dropouts within allocated treatment groups. Furthermore, a comparison of baseline characteristics of participants between study groups was made. Chi-square test was used for categorical variables and the Wilcoxon rank-sum test was used for ordinal variables.

Estimation of required sample size was based on a previous neuroimaging study to detect statistically significant results in functional MRI [30]. According to this, 20 participants in each group should be recruited; however, the goal was to recruit 30 participants in each group to compensate for dropouts. It was also considered to be sufficient to provide useful information about the feasibility of a future larger randomized controlled trial. A posthoc power analysis with characteristic pain intensity as the outcome measure was performed. The sample size required for a parametric test on the between-group comparison was computed, and 15% was added, as nonparametric testing could be used in a future randomized controlled trial [31].

All data management and statistical analyses were performed unblinded using StataSE (version 15.1; StataCorp LLC). A probability level of P<.05 was considered significant.

Participant flow through each stage of the trial is presented in Figure 2. The baseline characteristics of the sample are described in Table 2. Comparisons of baseline demographic, clinical, and psychosocial characteristics showed no statistical difference between treatment groups except that the occlusal splint group had a significantly greater number of jaw muscles with pain on palpation (P=.049; Table 2).

The between-group analysis of change in outcome measures (Table 3) showed no significant difference between the treatment groups at any follow-up regarding characteristic pain intensity (3 months: P=.58; 6 months: P=.41), pain-related disability (3 months: P=.51; 6 months: P=.12), jaw functional limitation (3 months: P=.45; 6 months: P=.90), degree of depression (3 months: P=.64; 6 months: P=.65), anxiety (3 months: P=.93; 6 months: P=.31), stress (3 months: P=.66; 6 months: P=.74), or catastrophizing (3 months: P=.86; 6 months: P=.85). Within-group analysis (Table 3) showed a significant reduction in jaw function limitation at the 6-month follow-up in the internet-based multimodal pain program group (Friedman: χ²₂=10.2, P=.04; Wilcoxon: z=–2.3, P=.02). In the occlusal splint group, characteristic pain intensity was significantly reduced at both 3- and 6-month follow-up (Friedman: χ²₂=25.1, P=.01; Wilcoxon 3 month: z=–3.0, P=.003; Wilcoxon 6 month: z=-3.3, P=.001) and jaw function limitation was reduced at the 6-month follow-up (Friedman: χ²₂=20.0, P=.045; Wilcoxon: z=–2.33, P=.02).

Comparison of pretreatment expectations on allocated treatment did not show any significant difference in positive (P=.56) or negative expectancy (P=.81) between groups (Table 2). In the internet-based multimodal pain program group there was a significant correlation between positive expectancy at baseline and change in pain-related disability (ρ=0.78, P=.04).

There was also a significant negative correlation between positive expectancy and change in the degree of depression (ρ=–0.78, P=.04). In the occlusal splint group, positive expectancy was positively correlated to change in catastrophizing (ρ=0.66, P=.01). No significant correlation was found between negative expectancy and change in treatment outcomes (Table 4).

An average of 101 patients were screened per month, and 5 patients per month went through the inclusion examination. From this, 2 participants per month were enrolled, and 43 participants were recruited during the a priori determined recruitment period—fewer than had initially been planned. An attempt was made to extend the recruitment period, but it was considered futile under the current circumstances. The attrition rate showed an unbalanced dropout between groups (Figure 3). The proportion of dropout at the 6-month follow-up was 65% (13/20) for the internet-based multimodal pain program group and 39% (9/23) for the occlusal splint group. In total, 51% (22/43) of the enrolled patients did not complete the 6-month follow-up. Figure 3 also shows how many participants completed each module in the internet-based multimodal pain program treatment. Of the 20 participants in the internet-based multimodal pain program, 14 started treatment, and 8 completed all 7 modules of the program. Of those who completed the program, the time spent between the start-up phone calls and the last follow-up phone call was a median of 14 (IQR 8-15) weeks. No adverse events were reported by participants randomized to internet-based multimodal pain program, and none choose to receive additional treatment at the 3-month follow-up.

The dropout analysis showed no difference between completers and dropouts in the internet-based multimodal pain program group regarding demographic factors, clinical or psychosocial characteristics (Multimedia Appendix 3). In the occlusal splint group, the dropouts were significantly younger (P=.02), had a lower proportion of married or de facto (P=.01), a lower proportion of participants with full time employment (P<.01), and a higher number of unspecific physical symptoms compared to completers (P=.03; Multimedia Appendix 3).

Posthoc power calculations showed a power of 11% and a sample size of 292 required for detecting a significant difference in characteristic pain intensity between 2 groups with α=.05 and β=.80 (characteristic pain intensity at the 6-month follow-up—internet-based multimodal pain program: mean 2.85; occlusal splint: mean 2.17; entire sample SD 1.92)

This study set out to assess the effectiveness of an internet-based multimodal pain program for adults with chronic TMD pain in comparison to occlusal splint therapy in general dentistry. No differences in clinical outcomes were found between the 2 treatment modalities. However, within the internet-based multimodal pain program group, jaw functional limitation was reduced and within the occlusal splint group, characteristic pain intensity and jaw functional limitation was reduced. As the study proceeded, a substantially higher dropout rate than expected was observed. Consequently, this pilot study showed that a randomized controlled trial with this design is not feasible.

This study did not find any difference in any of the recommended IMMPACT core outcome measures between the internet-based multimodal pain program and occlusal splint therapy. One possible explanation for this is the small sample size due to fewer participants recruited than planned and the substantial dropout rate. The lack of sufficient power due to the small sample size is a major limitation of this study regarding interpretation of treatment effect. Another possible explanation is our choice to include an active control condition. In this study, the effectiveness of internet-based multimodal pain program was compared to occlusal splint therapy—a treatment with moderate efficacy on TMD pain [5]. Our choice of this control condition in an early phase trial can be debated. However, according to the Helsinki declaration, “The benefits, risks, burdens, and effectiveness of a new intervention must be tested against those of the best proven intervention(s), except in the following circumstances: where no proven intervention exists, the use of placebo, or no intervention, is acceptable [32].”

Physical function is an important factor to assess in chronic pain studies [28]. In this study, we used the Jaw Function Limitation scale to assess jaw function. Within-group analysis indicated that the internet-based multimodal pain program increases jaw function for those with chronic TMD pain. This can be expected because the internet-based multimodal pain program comprises jaw exercises as well as psychoeducation and situational analyses that aim to reduce fear of movement [33]. The multimodal internet-based pain program therapy, therefore, seems to improve physical function. However, the findings cannot be generalized due to the small sample size. Potential treatment effects on the other outcome measures were not possible to determine in this study. The lack of change regarding pain intensity and pain-related disability was somewhat unexpected, as others have shown that multimodal treatment and cognitive behavior therapy are effective for patients with chronic orofacial pain [34-36], and internet interventions have been shown to improve pain interference and pain intensity [9,37].

The IMMPACT core outcome domain of participant ratings of global improvement was not assessed in this study since it is not a part of the DC/TMD instrument that was used in this study. Future studies should consider assessing global change because it could provide an estimate of overall change in symptoms in addition to those of the primary outcomes. This may be especially important when investigating a multicomponent treatment package where the relative contribution of each active component is unknown.

Within the occlusal splint group, characteristic pain intensity and jaw functional limitation were significantly reduced at the 6-month follow-up compared to baseline. The treatment effect of occlusal splint treatment on pain intensity was expected and is consistent with the literature [38]. A reduction in jaw functional limitation after occlusal splint therapy has also been reported [39]. It can, therefore, be assumed that the delivered control condition worked as expected.

IMMPACT recommends that expectations be considered as a phenotypic measure in clinical trials, and in this study, it was assessed with the multidimensional Stanford Expectations of Treatment Scale [40]. Research suggests that expectations influence the treatment outcome, and patients with positive expectations seem to benefit from medical treatment to a greater extent for a variety of conditions, including chronic pain [41]. The negative relation between positive expectancy and degree of depression in the internet-based multimodal pain program group is in line with this. Interestingly, this study also found a positive relation between positive expectancy and pain-related disability in the internet-based multimodal pain program group and that positive expectancy was positively associated with catastrophizing in the occlusal splint group. These relationships are interesting because it might suggest that positive expectations that are too high could have a negative impact on treatment outcome and also highlights the complex relationship of expectations on treatment and treatment outcome. Further studies are recommended to better understand the relationship between pretreatment expectations and outcomes of orofacial pain management protocols.

Individuals with chronic TMD pain were effectively recruited with this study protocol. All eligible patients except one chose to participate in the study, suggesting that participation was appealing. Retaining participants, on the other hand, turned out to be more of a challenge.

The high proportion of loss to follow-up confirms the difficulty of collecting data in longitudinal studies. However, the total proportion of dropouts in this trial (22/43, 51%) was relatively high compared to those for other clinical chronic pain trials, for which attrition ranged from 5% to 46% [42]. It was also surprising that some participants dropped out before the treatment start. A possible explanation for this might be the mandatory MRI examinations before and after the treatment period, particularly given that the MRI examinations were performed in a different city than that in which the patients were recruited. In addition, technical issues with the MRI scanner during a 6-month period in 2017 caused further inconvenience for the participants.

In addition to high attrition, unbalanced dropout for the 2 treatment modalities at the 6-month follow-up was observed. This imbalance could be explained neither by a difference between the treatment groups regarding baseline characteristics, treatment expectations, adverse effects, or the need of rescue treatment nor by differences in baseline characteristics between dropouts and completers. The dropout rate in the occlusal splint group was 39% (9/23). A systematic review of randomized controlled trials on occlusal treatments in TMD reported a dropout rate of less than 10% [43]. We believe that the MRI examinations were a major factor in the unexpectedly high dropout rate in this group. In the internet-based multimodal pain program group, the dropout rate of 65% is comparable to those of other internet-based trials; attrition has been recognized as one of the methodological challenges in the evaluation of eHealth apps [44], and future studies should consider an allocation process that compensates for this risk. It is, however, clear that that a randomized controlled trial with the current design is not feasible from economic and ethical perspectives due to the high attrition rate, even if the study may be considered practically feasible.

Further work is required to explore the potential of internet-based treatment for chronic orofacial pain. However, since the feasibility of this study design was suboptimal, different approaches have to be considered. For a future trial, adjustments in study design should be considered to reduce nonusage and dropout attrition. Inspired by a prior study [45] on an eHealth app for pain with low nonusage attrition, changes to the internet-based multimodal pain program could include optimization of scheme and triggers for reminders and offer different levels of clinical support according to the patient's preferences. Furthermore, the internet-based treatment used by that study had fewer modules that were less time consuming compared to those in this study's internet-based multimodal pain program [45]. Fewer and more focused modules should, therefore, be considered. Other possible changes to the internet-based multimodal pain program to make it more user-friendly include converting the offline workbook to a digital format and developing an app for mobile devices. In addition to improving the internet-based multimodal pain program itself, changes in the study protocol and design should be considered. We believe that the additional pre and posttreatment MRI examination was the primary reason for the high dropout rate in this study. Accordingly, a study of the treatment effect of internet-based multimodal pain program should not include MRI examinations. Also, conversion to a hybrid trial model with follow-up data collection online could make it more convenient to participate in the study and increase retention. Finally, novel study designs such as n-of-1 or noninferiority trials should be considered to optimize the quality and quantity of data acquired [46,47].

This study was not able to demonstrate a difference in treatment outcome between an internet-based multimodal pain program and occlusal splint therapy in patients with chronic TMD pain. However, within the internet-based multimodal pain program group, the results suggested that internet-based multimodal pain program improves jaw function. The results also confirmed the treatment effect of occlusal splint therapy in chronic TMD pain. Furthermore, the outcome of this pilot study showed that a randomized controlled trial with this design is not feasible due to a too high attrition rate.