JMIR Publications

ABSTRACT

Background:

Temporomandibular disorders (TMDs) are common chronic conditions involving orofacial pain and functional limitation. Digital therapeutics (DTx) have demonstrated efficacy in TMD management, yet the behavioral and clinical mechanisms underlying treatment response remain poorly characterized, particularly the question of whether behavioral modification or app engagement intensity drives therapeutic benefit.

Objective:

To investigate the behavioral mechanisms, responder profiles, and moderators of clinical response in a digital therapeutic intervention for TMD through a post hoc analysis integrating self-reported, server-derived, and clinician-rated outcome measures.

Methods:

We performed a post hoc secondary analysis of a multicenter, double-blind, sham-controlled randomized superiority trial conducted at two tertiary care centers in South Korea. The per-protocol cohort comprised 93 participants (DTx, n=44; sham, n=49). Five complementary analyses were applied: responder logistic regression at ≥30%, ≥50%, and ≥70% Visual Analog Scale (VAS) pain-reduction thresholds; subgroup comparison by Oral Behaviors Checklist (OBC) modifier status; causal mediation analysis using the potential outcomes framework with bootstrap confidence intervals; week-4 sensitivity analysis; and moderator analysis testing the Treatment × Patient Health Questionnaire-4 (PHQ-4) interaction on VAS change.

Results:

DTx assignment was consistently associated with clinically meaningful pain reduction across all three responder thresholds (adjusted odds ratios 3.21–5.39; all P<.05). Mediation analysis demonstrated that approximately 29.1% of the total VAS treatment effect was attributable to OBC-defined behavioral modification (Natural Indirect Effect −6.91 mm; 95% CI −13.23 to −0.58; P=.03), with the mediated proportion rising from 19.6% to 30.2% as responder thresholds became more stringent. OBC modifiers achieved substantially greater pain reduction than non-modifiers (−45.71 vs −22.61 mm; difference −23.11; 95% CI −36.97 to −9.24; P<.01) despite no significant differences across any objective app engagement metric, indicating that behavioral adaptation rather than usage volume drives therapeutic benefit. Treatment odds ratios were 33–42% higher at week 4 than at the 6-week endpoint (week-4 ORs 4.48–7.15), identifying this timepoint as a clinically actionable decision node. Baseline psychological distress (PHQ-4 ≥3) significantly moderated treatment response (interaction β=−19.63; SE=9.17; P=.04).

Conclusions:

Behavioral modification — not app engagement volume — is the operative mechanism of DTx efficacy in TMD, accounting for approximately 29% of the total pain-reduction effect. The dissociation between behavioral adaptation and engagement metrics challenges the use of usage volume as a surrogate for therapeutic dose. Week 4 emerges as an early decision node for adaptive treatment protocols, and baseline psychological distress may inform precision patient selection. These findings support a mechanism-informed, precision-oriented approach to digital therapeutic design and deployment in TMD management. Clinical Trial: Clinical Research Information Service KCT0009493; https://cris.nih.go.kr/cris/search/detailSearch.do