Prognostic factors for pain intensity and function in individuals with temporomandibular disorders (TMDs): a systematic review and meta-analysis

Abstract

Objectives

To investigate the prognosis and prognostic factors associated with pain intensity and function in individuals with temporomandibular disorders (TMDs). Secondary objectives included identifying prognostic factors related to symptom progression, subsequent treatment needs and psychosocial outcomes.

Design

Systematic review with meta-analysis.

Data sources

Medline, Dentistry and Oral Sciences Source, SportDiscus and CINAHL (EBSCOhost) were searched up to September 2025.

Eligibility criteria

Prospective longitudinal studies examining prognostic factors for pain intensity or functional outcomes in adults with TMD, with a minimum follow-up of 3 months.

Data extraction and synthesis

Two independent reviewers screened titles and abstracts and assessed full texts for eligibility. Data were extracted in duplicate using a standardised protocol that included study characteristics, prognostic factors and outcomes. Risk of bias was assessed using the QUIPS (Quality in Prognosis Studies) tool, and the certainty of evidence was graded using a modified GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach. Meta-analyses were performed with random-effects models, reporting ORs and 95% CIs. Heterogeneity was evaluated using I² (I-squared) and Tau² (Tau-squared) statistics. Where pooling was not feasible, findings were synthesised narratively.

Results

Nine prospective studies met the inclusion criteria; eight were included in quantitative analyses (n=4282) and one large cohort was synthesised narratively (n=94 769), for a total of 99 051 participants. After standardising definitions, 56 candidate factors were analysed. In adjusted models for pain, pain provoked by movement or palpation was associated with a worse course (eg, joint pain with sound: OR=2.10, 95% CI 1.39 to 3.18; muscle pain during movement: OR=2.10, 95% CI 1.33 to 3.31), while shorter pain duration (OR=0.25, 95% CI 0.09 to 0.70) and greater pain-free mouth opening (OR=0.60, 95% CI 0.40 to 0.90) were protective. For the function outcome, pain intensity was associated with poorer outcomes (OR=1.39, 95% CI 1.14 to 1.69), whereas age, sex, depression, somatisation, disability days and self-efficacy showed no consistent associations.

Conclusions

In adults with TMD, pain provoked by movement or palpation and higher pain intensity were consistently associated with less favourable prognoses. Conversely, shorter pain duration and greater pain-free opening were associated with better outcomes. These associations are non-causal and based on low to very low certainty evidence amid methodological heterogeneity. While they may inform risk stratification, they should not guide treatment decisions without confirmatory longitudinal studies using standardised outcomes and improved control of confounders.

PROSPERO registration number

CRD42024557159.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Prospectively registered protocol on Prospective Register of Systematic Reviews and Open Science Framework, and conducted aligned with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses), with prespecified eligibility criteria and outcomes.

• Comprehensive multidatabase search to September 2025, with independent duplicate screening and data extraction using a standardised form.

• Risk of bias appraised with Quality in Prognosis Studies (QUIPS) and certainty of evidence graded with an adapted Grading of Recommendations Assessment, Development and Evaluation framework for prognostic factor research.

• Standardised definitions of prognostic factors; prioritised adjusted estimates; and random-effects meta-analyses with I² (I-squared)/Tau² (Tau-squared) where pooling was feasible.

• Limitations include heterogeneity and moderate to high risk of bias among included studies; however, methodological transparency and use of adjusted estimates and QUIPS enhanced the reliability of findings.

Introduction

Temporomandibular disorders (TMD) are a group of musculoskeletal conditions that affect the temporomandibular joint (TMJ), masticatory muscles and associated structures.^{1 2} The aetiology of TMD is complex and multifactorial, and its clinical course varies considerably among individuals.³,⁷ While some individuals recover spontaneously or with minimal intervention, others develop persistent symptoms, including chronic pain and functional impairment.⁸

Chronic pain related to TMD can substantially affect quality of life and psychological well-being.^{9 10} Consequently, there is growing interest in identifying prognostic factors that may influence the course of TMD. Prognostic factors are characteristics associated with future health outcomes, regardless of the treatment provided. Identifying such factors can support clinical decision-making, facilitate personalised care and enhance the allocation of healthcare resources.¹¹,¹³

Although some studies have examined potential prognostic indicators—such as pain intensity, psychological distress and parafunctional behaviours—the available evidence remains fragmented and methodologically inconsistent.^{14 15} Moreover, no systematic review has yet synthesised the existing evidence on prognostic factors related to pain intensity and functional outcomes in individuals with TMD.

TMD is a highly prevalent condition with a substantial impact on quality of life, yet prognostic uncertainty continues to hinder effective long-term management. To address this gap, this systematic review synthesised evidence from longitudinal observational designs (prospective cohort studies), which are appropriate for prognosis research, to identify prognostic factors associated with pain intensity and functional outcomes in individuals with TMD across short- and long-term follow-up.

Methods

Study design

This study is a systematic review and meta-analysis conducted in accordance with the MOOSE (Meta-analysis of Observational Studies in Epidemiology) and PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, with the respective checklists provided in online supplemental appendix A (MOOSE) and online supplemental appendix B (PRISMA)¹⁶ and with reference to the Cochrane Handbook for Systematic Reviews of Interventions,¹⁷ with adaptations appropriate for prognostic research. The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO (CRD42024557159): https://www.crd.york.ac.uk/PROSPERO/view/CRD42024557159) and published on the Open Science Framework (https://doi.org/10.17605/OSF.IO/U3VXR).

Eligibility criteria

Types of studies

We included prospective and retrospective observational studies that examined the prognosis and prognostic factors of TMD. Case–control studies and randomised controlled trials that met the inclusion criteria and aligned with the study objectives were also considered. All studies were independently assessed for eligibility by two reviewers.

To be eligible, studies were required to include a baseline assessment (defined as the start of the study) of participant characteristics and a follow-up period classified as either short-term (≤3 months) or long-term (>3 months).¹⁸

Studies were categorised based on their stated objectives, study design and analytical approach. These characteristics were also considered when evaluating the strength of the evidence.

Population

We included studies involving individuals of any age or sex diagnosed with TMD—including articular, myogenic, arthrogenic or mixed types—at either acute (≤3 months) or chronic (>3 months) stages, based on clinical signs and symptoms or imaging findings.

We excluded studies involving individuals who had undergone non-elective surgery (eg, for facial trauma or oncological conditions), those with specific comorbidities (eg, fractures, cancer or neurological disorders), as well as case reports, cadaveric studies and trials lacking a relevant prognostic focus.

Outcomes

The primary outcome was the longitudinal change in pain intensity and functional status in individuals with TMD. Secondary outcomes included prognostic factors associated with symptom progression, subsequent treatment requirements and psychosocial consequences.

Review process

The review process followed three structured phases: Planning—definition of databases, development and pilot-testing of the search strategy, identification and selection—electronic and manual searches, screening, risk of bias assessment and data extraction, and analysis and synthesis—data interpretation, subgroup analyses, evidence grading and preparation of the final report. This phased approach was designed to maximise methodological rigour and minimise bias.

All search results were exported to Rayyan software, where duplicates were removed. Two reviewing researchers (MA and BML) independently and blinded screened titles and abstracts, followed by full-text assessment against the eligibility criteria. Disagreements were resolved by consensus or adjudication with a third reviewer (LF). In the case of multiple publications reporting overlapping data, the most complete report was selected as the primary source.

Information sources and search strategy

We systematically searched the following electronic databases: PubMed/MEDLINE, Embase (Elsevier), Dentistry and Oral Sciences Source, SportDiscus and CINAHL (EBSCOhost). In addition, the reference lists of included articles were screened to identify any potentially relevant studies. The searches covered the period from database inception to 16 September 2025, the date of the last update. No restrictions were applied regarding language or publication status.

Comprehensive search strategies were developed by experienced researchers and adapted for each database. The strategies combined terms related to prognosis and prognostic factors with those referring to TMD, using Boolean operators and database-specific indexing terms (eg, Medical Subject Headings). The complete search strategies for each database are provided in online supplemental appendix C.

To enhance the transparency and reliability of the search process, we applied additional strategies. The selection of databases was deliberately limited to those most relevant and specific to the research question, aiming to maximise the precision of the retrieved results. Grey literature searches were not performed, thereby reducing potential sources of confounding. Manual searches of reference lists of relevant systematic reviews were conducted to identify potentially eligible full-text articles.

Procedures for contacting authors

In cases of missing data or insufficient information regarding effect sizes, study authors were contacted via email. Two attempts were made to retrieve missing data, and both the need for contact and the outcomes of these attempts were documented.

Data extraction

Data extraction was performed independently by two reviewers (MA and MGMC) using a piloted, standardised form. Training and calibration exercises were conducted before formal extraction to ensure consistency. Extracted data included: author, year, country, diagnostic criteria, study design, recruitment period, sample size, participant characteristics (age, sex distribution, comorbidities), prognostic factors, outcomes, follow-up duration, attrition and statistical models.

Discrepancies were resolved by consensus or adjudication with a third reviewer (LF). A proportion of studies was cross-checked for consistency. Data management was conducted using Excel spreadsheets designed explicitly for this review.

Risk of bias

Risk of bias was assessed independently by two reviewers (MA and LF) using the QUIPS tool (Quality in Prognosis Studies).¹⁹ Six domains were considered: study participation, attrition, measurement of prognostic factors, measurement of outcomes, study confounding and statistical analysis/reporting. Each domain was rated as low, moderate or high risk of bias. Disagreements were resolved by consensus or adjudication by a third reviewer.

Quality of evidence

The certainty of the evidence was appraised using a modified GRADE (Grading of Recommendations Assessment, Development and Evaluation) framework adapted for prognostic research.²⁰ Certainty was rated as high, moderate, low or very low, considering the phase of investigation (exploratory vs confirmatory), study limitations, heterogeneity, precision, effect size, generalisability and potential reporting bias. This approach followed the adaptation previously outlined by Huguet et al and subsequently applied in prognostic factor reviews.

Two independent reviewers (MA and LF) conducted all certainty assessments. In cases of disagreement, consensus was reached through discussion; when necessary, a third reviewer (BML) was consulted to resolve discrepancies.

Statistical analysis

Where sufficient data were available, we performed random-effects meta-analyses (Comprehensive Meta-Analysis software) to account for study heterogeneity between. ORs were the primary effect measure; risk ratios (RRs),¹⁷ risk differences and regression coefficients were converted to log ORs when necessary. Standard errors were calculated from CIs. Both adjusted and unadjusted estimates were analysed.

Heterogeneity was assessed by visual inspection of forest plots and quantified using I² (I-squared) and Tau² (Tau-squared). Given the small number of contributing studies, I² was interpreted cautiously (especially when k<5) and not computed/interpreted when k<2.^{21 22}

Slight study/publication bias was not assessed using funnel plots or Egger’s/Begg’s tests because fewer than 10 studies contributed to any meta-analysis (per Cochrane guidance); potential reporting bias was considered within the adapted GRADE framework. When pooling was infeasible or interpretability was limited, clinical heterogeneity was examined narratively across population characteristics, outcome measures, follow-up duration and study design. Subgroup analyses and meta-regression were planned when study numbers permitted, and sensitivity analyses tested robustness by excluding high risk-of-bias studies and comparing effect measures (eg, ORs vs RRs).

Potential publication bias was addressed by contacting study authors at least twice to obtain missing data; no imputation procedures were applied. Narrative synthesis was undertaken when quantitative pooling was not feasible.

Patient and public involvement

This research was done without patient or public involvement. Patients or members of the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Results

Study Selection

The initial search yielded 13 258 records. After removing duplicates, 11 822 articles remained for title and abstract screening. Of these, 39 were selected for full-text review, and 9 met the inclusion criteria. Among the 27 articles excluded at this stage, 9 were not prognostic studies, 8 reported inappropriate outcomes, 7 were inaccessible despite author contact attempts and 6 lacked sufficient methodological detail for inclusion (figure 1).

Figure 1. PRISMA 2009 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow diagram of study selection. Flow diagram illustrating the number of records identified, screened, excluded, assessed for eligibility and included in the review, with reasons for exclusion at the full-text stage.

One of the included studies did not provide sufficient clarity regarding data for specific prognostic factors. The corresponding author was contacted for clarification; however, no response was received. As a result, the affected variable was excluded to preserve the integrity of the data analysis.

Characteristics of included studies

All included studies employed a prospective longitudinal design, comprising six cohort studies, one case–control study and two studies described only as prospective longitudinal. The investigations were conducted between 2003 and 2019, with one based in Sweden, three in Finland, four in the USA and two in other countries (South Korea and the Netherlands). The total sample comprised 99 051 participants (50 065 women and 48 986 men), indicating a nearly balanced sex distribution. The proportion of male participants was approximately 1.4% higher than that of females, which is unlikely to have influenced the overall representativeness of the cohort. Participants’ ages ranged from 18 to 82 years. Although secondary outcomes were predefined, they could not be analysed due to heterogeneity and inconsistent reporting across studies. The characteristics of included studies are summarised in table 1.

Table 1. Characteristics of included studies.

Author, Year	Country	Study design	Sample size (n)	Follow-up	Diagnostic criteria	Outcome(s)	Candidate prognostic variables
Banafa, 202027	Finland	Prospective longitudinal	1210	11 years	Clinical examination: MMO, TMJ auscultation, TMJ/MM palpation	TMD pain	Gender; age; number of teeth; denture status; educational level; masticatory muscle pain on palpation; TMJ pain on palpation
Banafa et al, 202328	Finland	Prospective longitudinal	1087	11 years	Clinical examination: MMO, TMJ auscultation, TMJ/MM palpation	TMD pain	Cynical hostility; somatisation
Forssell, 201723	Finland	Prospective longitudinal	399	1 year	RDC/TMD	Pain and function	Gender; level of education; age; parafunctions; time since onset; CPI; pain-related disability; number of disability days; functional jaw limitations; SCL-90 depression; SCL-90 somatisation; sleep dysfunction; pain-related worry; anxiety; stress; catastrophising; pain control; ability to decrease pain; perceived risk of chronicity; healthcare visits; other pain conditions; RAND-36 physical function
Jung, 202126	South Korea	Prospective longitudinal	486	2.8 years	Clinical TMD diagnosis	TMD-related pain	Chronic pain; muscle tenderness; somatisation; psychoticism
Lövgren et al, 202525	Sweden	Prospective longitudinal	94 769	7 years	3Q/TMD screening	TMD pain/jaw locking	Sex; age
Meloto et al, 201914	USA	Prospective case-control	147	8 months	DC/TMD	TMD persistence and function	Pain-free opening; maximal opening; provocation pain; joint sounds; CPI; pain interference; chronic pain grade; functional limitation; depression; anxiety; parafunction
Rammelsberg et al, 200313	USA	Prospective longitudinal	368	5 years	RDC/TMD	TMD pain	Age; sex; depression; somatisation without pain; pain intensity; pain frequency; number of intraoral sites; number of extraoral sites; number of body pain sites
Rollman, 201324	Netherlands	Prospective longitudinal	129	6 months	Self-report+clinical confirmation	Pain and function	Pain duration; number of practitioners consulted; functional hindrance; pain coping; pain-related disability
Slade et al, 201447	USA	Prospective longitudinal	456	6 months	Clinical history+palpation examination	Pain and function	Pressure pain thresholds

Study-level descriptors are summarised for each included article: country, study design, sample size (n), length of follow-up, diagnostic criteria/assessment, primary outcome(s) (pain and/or function) and candidate prognostic variables. Candidate variables listed reflect those extracted from each study and do not imply statistical significance in adjusted analyses. See online supplemental table H and I for a crosswalk of outcome instruments (eg, VAS/NRS for pain; JFLS/MFIQ/self-efficacy for function), including scale ranges and time points.

CPI, characteristic pain intensity; DC/TMD, diagnostic criteria for temporomandibular disorders; JFLS, Jaw Functional Limitation Scale; MFIQ, Mandibular Function Impairment Questionnaire; MM, masticatory muscles; MMO, maximum mouth opening; NRS, Numerical Rating Scale; 3Q/TMD, Three-question TMD screening tool; RDC/TMD, research diagnostic criteria for temporomandibular disorders; SCL-90, Symptom Checklist-90; TMD, temporomandibular disorder; TMJ, temporomandibular joint; VAS, Visual Analogue Scale.

Statistical analysis

Across the eight included studies, 80 prognostic factors were initially identified. This number was subsequently refined to 56 for analysis due to the merging of similarly defined variables labelled differently across studies, the exclusion of variables with unclear measurement methods and misalignment with the specified outcomes.

In the adjusted analyses, 38 prognostic factors were evaluated for pain. Of these, 9 were identified as potentially contributing to a worse clinical course of TMD, 27 were classified as non-significant and 2 emerged as protective factors (table 2). A meta-analysis was performed on four studies,^{13 14 23 24} encompassing seven variables: gender, age, depression, somatisation level, number of disability days, pain intensity and self-efficacy (online supplemental D).

Table 2. Prognostic factors related to pain in patients with TMDs. Adjusted analysis.

Prognostic factor	Direction	Adjusted effect (OR, 95% CI)
Age	Non-significant	1.00 (0.98 to 1.02)
Anxiety	Non-significant	0.90 (0.70 to 1.16)
Arthralgia/arthritis/arthrosis	Non-significant	0.97 (0.28 to 3.33)
Chewing limitation	Non-significant	1.20 (0.80 to 1.80)
Click	Non-significant	1.10 (0.71 to 1.71)
Cynical hostility	Non-significant	1.00 (0.95 to 1.05)
Depression	Non-significant	1.06 (0.81 to 1.39)
Disc displacement	Non-significant	2.01 (0.71 to 5.66)
Disability	Non-significant	0.99 (0.98 to 1.00)
Duration of pain	Protective	0.25 (0.09 to 0.70)
Free opening	Protective	0.60 (0.40 to 0.90)
Gender	Non-significant	1.12 (0.47 to 2.65)
Headache—familiar on palpation	Worse prognosis	1.50 (1.10 to 2.05)
Headache—familiar with mobility	Non-significant	1.20 (0.80 to 1.80)
Jaw noises—any sound	Non-significant	1.20 (0.90 to 1.60)
Joint pain accompanying sound	Worse prognosis	2.10 (1.38 to 3.17)
Maximum assisted opening	Non-significant	0.80 (0.58 to 1.10)
Maximum opening	Protective	0.70 (0.49 to 0.99)
Muscle pain in mobility	Worse prognosis	2.10 (1.33 to 3.31)
Muscle pain on palpation	Worse prognosis	1.75 (1.18 to 2.60)
Number of body pain sites	Worse prognosis	1.97 (1.11 to 3.50)
Number of disability days	Non-significant	1.08 (0.94 to 1.24)
Number of extraoral sites	Worse prognosis	1.18 (1.03 to 1.35)
Number of healthcare visits	Worse prognosis	1.19 (1.02 to 1.39)
Number of intraoral sites	Non-significant	0.92 (0.61 to 1.38)
Number of other pain conditions	Non-significant	1.31 (0.98 to 1.75)
Opening limitation	Non-significant	0.90 (0.60 to 1.35)
Parafunction	Non-significant	1.10 (0.80 to 1.51)
Pressure pain thresholds	Worse prognosis	1.36 (1.11 to 1.66)
Self-efficacy	Non-significant	0.97 (0.59 to 1.59)
Somatisation level	Non-significant	1.47 (0.52 to 4.20)
Prognostic factor	Protective factors	Non-significant factors

Adjusted associations between candidate prognostic factors and pain outcome in individuals with TMDs. ‘Protective’ indicates OR <1 with 95% CI not crossing 1; ‘Worse prognosis’ indicates OR >1 with 95% CI not crossing 1; ‘non-significant’ means 95% CI including 1. Estimates reflect the most fully adjusted model available in each primary study.

TMDs, temporomandibular disorders.

In the adjusted analysis for function, 29 factors were evaluated; of these, 3 showed potential protective effects (maximum opening, pain duration and pain-free opening), and 8 were identified as risk factors (table 3). A meta-analysis was performed on four studies^{13 14 23 24} (online supplemental E).

Table 3. Prognostic factors related to functional prognosis in patients with TMDs. Adjusted analysis.

Prognostic factor	Direction	Adjusted effect (OR, 95% CI)
Age*	Non-significant	1.00 (0.98 to 1.02)
Anxiety	Non-significant	0.90 (0.70 to 1.16)
Any sound	Non-significant	1.20 (0.90 to 1.60)
Chewing limitation	Non-significant	1.20 (0.80 to 1.80)
Click	Non-significant	1.10 (0.71 to 1.71)
Crepitus	Non-significant	1.00 (0.70 to 1.43)
Depression*	Non-significant	1.10 (0.80 to 1.51)
Disability	Non-significant	0.99 (0.98 to 1.00)
Familiar headache on palpation	Worse prognosis	1.50 (1.10 to 2.05)
Familiar headache w/mobility	Non-significant	1.20 (0.80 to 1.80)
Familiar joint pain accompanying sound	Worse prognosis	1.80 (1.30 to 2.49)
Gender*	Non-significant	1.28 (0.44 to 3.71)
Joint accompanying sound	Worse prognosis	2.10 (1.39 to 3.17)
Maximal assisted opening	Non-significant	0.80 (0.58 to 1.10)
Maximum opening	Protective	0.70 (0.49 to 0.99)
Muscle pain in mobility	Worse prognosis	2.10 (1.33 to 3.31)
Muscle pain on palpation	Worse prognosis	1.75 (1.18 to 2.60)
Number of disability days*	Non-significant	1.08 (0.98 to 1.84)
Number of healthcare visits	Worse prognosis	1.19 (1.02 to 1.34)
Number of pain conditions	Non-significant	1.31 (0.983 to 1.84)
Opening limitation	Non-significant	0.90 (0.60 to 1.35)
Pain duration	Protective	0.25 (0.09 to 0.70)
Pain-free opening	Protective	0.60 (0.40 to 0.90)
Pain intensity*	Worse prognosis	1.39 (1.14 to 1.69)
Parafunction	Non-significant	1.10 (0.80 to 1.51)
Self-efficacy*	Non-significant	0.97 (0.59 to 1.59)
TMJ pain in mobility	Non-significant	1.50 (1.00 to 2.25)
TMJ pain on palpation	Worse prognosis	1.51 (1.09 to 2.10)
Verbal/emotional expression limitation	Non-significant	0.90 (0.59 to 1.37)

Adjusted associations between candidate prognostic factors and functional outcome in individuals with TMDs. ‘Protective’ indicates OR <1 with 95% CI not crossing 1; ‘Worse prognosis’ indicates OR >1 with 95% CI not crossing 1; ‘non-significant’ indicates 95% CI including 1. Estimates reflect the most fully adjusted model available in each primary study.

(*) were assessed in more than one study

TMDs, temporomandibular disorders; TMJ, temporomandibular joint.

The unadjusted analyses were conducted exclusively for pain outcomes, as studies evaluating functional outcomes predominantly reported adjusted models or used heterogeneous continuous measures (eg, JFLS (Jaw Functional Limitation Scale), MFIQ (Mandibular Function Impairment Questionnaire), self-efficacy scales) that precluded standardisation of unadjusted estimates. The unadjusted analyses yielded different results: 38 prognostic factors were examined, of which 17 were potentially associated with a worse clinical course, 20 showed no significant association and 1 was identified as a protective factor. Complete unadjusted effect estimates and corresponding CIs are provided in online supplemental F. In this analysis, meta-analyses were feasible for a larger number of variables, including five non-significant factors (age, depression, somatisation level, disability and function). Quantitative synthesis for functional outcomes was therefore restricted to adjusted analyses.

Prognosis

Although five studies addressed the prognosis of TMDs, it was not possible to perform a quantitative analysis of the natural course of the condition. These studies primarily examined how specific factors—such as psychosocial variables, baseline pain intensity and symptom duration—influenced the likelihood of persistent pain or functional deterioration. Notably, the study by Lövgren et al²⁵ employed a Markov multistate model to evaluate transitions between clinical states over time; however, its primary focus remained on identifying prognostic factors (eg, sex, symptom duration) rather than describing the spontaneous clinical evolution of TMD.

To be classified as an actual natural course study, an investigation would need to follow participants longitudinally without focusing on predictive variables, reporting purely descriptive data such as rates of improvement, worsening or stability in pain and function, rather than attempting to explain these changes through explanatory models.

Prognostic factors for pain and function

Poor prognostic factors for pain

Nine factors were classified as potentially contributing to a poorer prognosis in individuals with TMD: familiar headache on palpation (OR=1.50, 95% CI 1.10 to 2.05), familiar joint pain accompanying sound (OR=1.80, 95% CI 1.30 to 2.49), joint pain accompanying sound (OR=2.10, 95% CI 1.39 to 3.18), muscle pain during movement (OR=2.10, 95% CI 1.33 to 3.31), muscle pain on palpation (OR=1.75, 95% CI 1.18 to 2.60), number of painful body sites (OR=1.97, 95% CI 1.11 to 3.50), pain frequency (OR=1.72, 95% CI 1.15 to 2.58), pressure pain thresholds (OR=1.36, 95% CI 1.11 to 1.66) and TMJ pain on palpation (OR=1.50, 95% CI 1.09 to 2.10) (table 2). Among these, only joint pain accompanied by noise and muscle pain during movement showed a strong association with the outcome. The remaining variables demonstrated weak to moderate associations, requiring cautious interpretation of the results.

In addition, Lövgren et al²⁵ applied a Markov multistate model to investigate transitions between clinical states of TMD pain, capturing the probabilities of remission, recurrence, persistence and symptom absence over time. The study included 94 769 individuals (49.9% women) and revealed that women had a higher likelihood of developing and maintaining pain, as well as a lower probability of recovery, compared with men. Longer symptom duration was also associated with an increased risk of persistence of pain. Although HRs were reported instead of ORs and thus were not included in the meta-analysis, these findings qualitatively corroborate the overall pattern of poorer pain prognosis among females and individuals with chronic or long-standing TMD symptoms.

Poor prognostic factors for function

A total of 29 prognostic variables were examined across studies evaluating functional outcomes. 18 showed no significant association with function, 3 were considered potentially protective—maximal mouth opening (OR=0.70, 95% CI 0.49 to 0.95), pain duration (OR=0.25, 95% CI 0.09 to 0.70) and pain-free opening (OR=0.60, 95% CI 0.40 to 0.90)—and 8 were identified as possible risk factors: familiar headache on palpation (OR=1.50, 95% CI 1.10 to 2.05), familiar joint pain accompanying sound (OR=1.80, 95% CI 1.30 to 2.49), joint pain accompanying sound (OR=2.10, 95% CI 1.39 to 3.18), muscle pain during movement (OR=2.10, 95% CI 1.33 to 3.31), muscle pain on palpation (OR=1.75, 95% CI 1.18 to 2.60), number of healthcare visits (OR=1.19, 95% CI 1.02 to 1.39), pain intensity (OR=1.39, 95% CI 1.14 to 1.69) and TMJ pain on palpation (OR=1.51, 95% CI 1.09 to 2.10). Of these, only three variables were eligible for meta-analysis: pain intensity (OR=1.39, 95% CI 1.14 to 1.69), number of disability days (OR=1.08, 95% CI 0.98 to 1.84) and self-efficacy (OR=0.97, 95% CI 0.59 to 1.59). Pain intensity was significantly associated with worse functional outcomes, whereas self-efficacy and the number of disability days were not. The heterogeneity in the measurement of functional outcomes (eg, JFLS, MFIQ and self-efficacy scales) likely contributed to the observed variability in effects (table 3).

Sensitivity analysis and subgroups

Prespecified sensitivity analyses—excluding high risk-of-bias studies and prioritising adjusted over unadjusted estimates—did not change the direction of the pooled effects; in a few cases, CIs widened/narrowed modestly. Leave-one-out analyses were not undertaken because no pooled factor included ≥3 studies (k<3). Subgroup analyses were not conducted due to insufficient studies per subgroup (k≤2), as specified in the statistical analysis plan.

Non-significant prognostic factors

27 variables were found to be non-significant for pain. Among these, seven were included in the meta-analysis: gender, age, depression, somatisation level, number of disability days, pain intensity and self-efficacy (table 2). All of these demonstrated weak to moderate associations with the outcome.

Regarding functional outcomes, 18 prognostic variables showed no significant association with mandibular function. These included demographic, psychosocial and pain-related factors such as age, anxiety, depression, somatisation and sleep quality. In the quantitative synthesis, two of the three meta-analysed variables—self-efficacy (OR=0.97, 95% CI 0.59 to 1.59) and number of disability days (OR=1.08, 95% CI 0.98 to 1.84)—were also not statistically significant, indicating limited predictive value for functional impairment (table 3).

Protective factors

Two factors were identified as potentially protective relative to pain: pain duration (OR=0.25, 95% CI 0.089 to 0.702) and pain-free mouth opening amplitude (OR=0.60, 95% CI 0.400 to 0900) (table 2). Meta-analysis was not feasible for these variables, as each was investigated in only one study. Both factors exhibited wide CIs, which limited the precision of the findings. Pain duration showed an OR <0.5, suggesting a strong protective effect, whereas pain-free mouth opening was associated with a moderate protective effect (table 2, online supplemental D).

In relation to functional outcomes, three variables were identified as potentially protective: maximal mouth opening (OR=0.70, 95% CI 0.49 to 0.95), pain duration (OR=0.25, 95% CI 0.09 to 0.70) and pain-free opening (OR=0.60, 95% CI 0.40 to 0.90). All three demonstrated significant associations, indicating that greater mandibular mobility and shorter pain duration were associated with a better functional prognosis. As each was analysed in a single study, meta-analysis was not feasible and the results should be interpreted with caution (table 3, online supplemental E).

Risk of bias

The risk of bias assessment indicated that study attrition was the most compromised domain, with four studies rated as high risk^{14 23 26 27} and four as moderate risk.^{13 24 25 28} Study confounding also emerged as a critical concern, with two studies rated as high risk^{13 24} and six as moderate risk,^{1423 25},²⁸ as summarised in table 4.

Table 4. QUIPS tool.

Study (Author, Year)	Study participation	Study attrition	Prognostic factor measurement	Outcome measurement	Study confounding	Statistical analysis and presentation	Overall risk of bias
Banafa et al, 202027	Low RoB	High RoB	Low RoB	Low RoB	Moderate RoB	Low RoB	Moderate
Banafa et al, 202328	Low RoB	Moderate RoB	Low RoB	Low RoB	Moderate RoB	Low RoB	Moderate
Forssell, 201723	Moderate RoB	High RoB	Low RoB	Low RoB	Moderate RoB	Low RoB	Moderate–high
Jung, 202126	Low RoB	High RoB	Low RoB	Low RoB	Moderate RoB	Low RoB	Moderate
Lövgren et al, 202525	Low RoB	Moderate RoB	Low RoB	Moderate RoB	Moderate RoB	Low RoB	Moderate
Meloto et al, 201914	Low RoB	High RoB	Low RoB	Low RoB	Moderate RoB	Low RoB	Moderate
Rammelsberg et al, 200313	Low RoB	Moderate RoB	Low RoB	Low RoB	High RoB	Low RoB	Moderate
Rollman et al, 201324	Low RoB	Moderate RoB	Low RoB	Low RoB	High RoB	Low RoB	Moderate–high
Slade et al, 201447	Low RoB	Low RoB	Low RoB	Low RoB	Low RoB	Low RoB	Low

RoB was evaluated across six domains: (1) study participation, (2) study attrition, (3) prognostic factor measurement, (4) outcome measurement, (5) study confounding and (6) statistical analysis and reporting. Each domain was rated as low, moderate or high risk of bias according to QUIPS guidance. Overall RoB was qualitatively judged based on the balance across domains, giving greater weight to confounding and outcome measurement, which were considered the most influential domains for prognostic validity (Hayden et al).⁴⁸

QUIPS, Quality in Prognosis Studies; RoB, Risk of Bias.

In contrast, most studies demonstrated low risk of bias across the domains of study participation, prognostic factor measurement, outcome measurement and statistical analysis and presentation, strengthening confidence in the reliability of findings in these areas.

Overall, most studies were rated as having moderate to moderately high risk of bias, underscoring the need for more rigorous methodological strategies to reduce attrition and improve control for confounding factors in future prognostic research (table 4).

Modified version of GRADE for prognostic factors study

Although several studies adequately met criteria for study participation, prognostic factor measurement and outcome measurement, the overall risk of bias was rated as moderate to high for most prognostic factors, mainly due to inadequate control of confounding and incomplete follow-up. These findings highlight the need for more rigorous methodological strategies, including better management of missing data and improved adjustment for confounders, to strengthen confidence in future prognostic research (online supplemental G).²⁰

Study limitations

Only one study was rated as having a low risk of bias, while five were rated as moderate and two as moderately high. This represents a significant limitation, particularly affecting the quality of evidence in the domains of attrition and confounding (online supplemental G).

Inconsistency

Although some variation in effect estimates was observed, most studies reported effects in the same direction and of comparable magnitude. Nevertheless, wide CIs frequently crossed the null line, suggesting statistical uncertainty. Variables exhibiting consistent associations included age, headache on palpation, joint sound, joint pain with sound, muscle pain during movement or palpation, number of painful body sites, pain frequency, pressure pain thresholds, TMJ pain during movement or palpation and limitations in verbal and emotional expression (online supplemental G).

Indirectness

Overall, the included studies were well aligned with the target population, prognostic factors, outcomes, clinical context and follow-up period defined in the review question. This indicates a low level of indirectness across studies.

Imprecision

Imprecision was assessed based on the width of CIs and sample sizes. Most factors demonstrated considerable imprecision, with wide CIs often crossing the null line, which reduced the certainty of the evidence. Only nine prognostic factors—age, headache on palpation, joint sound, joint pain with sound, muscle pain during movement, muscle pain on palpation, number of painful body sites, pain frequency and pressure pain thresholds—showed adequate precision to support more robust evidence (online supplemental G).

Publication bias

Only six prognostic factors were investigated in more than one study, and all relevant studies were rated as having moderate to high risk of bias. Furthermore, none of the included studies were confirmatory in design. This limited the ability to assess potential publication bias, although selective reporting remains possible.

Discussion

This systematic review synthesised prognostic evidence for pain and function in TMD. Nine prospective studies met the inclusion criteria, and 56 candidate variables were analysed after standardising definitions. In adjusted pain models, two factors appeared protective (shorter pain duration; greater pain-free opening) and nine pain-related clinical signs—chiefly pain provoked by movement or palpation—were associated with a poorer course; most other variables showed non-significant or small effects. For function, three factors (maximal mouth opening, shorter pain duration, pain-free opening) were protective, and eight showed risk associations. Only three variables were meta-analysed: pain intensity was associated with worse function, whereas self-efficacy and number of disability days were not statistically significant.

Although five studies addressed prognosis, none qualified as actual natural-course investigations focused on descriptive trajectories. Most used multivariable models to identify predictors, not spontaneous evolution. Lövgren et al²⁵ used a Markov multistate model. They found that women were more likely to develop and maintain pain, less likely to recover, and that longer symptom duration increased persistence. Because HRs (not ORs) were reported, this study was excluded from the quantitative synthesis.

Nevertheless, its findings qualitatively support the overall pattern of a less favourable pain prognosis among women and individuals with longer-lasting symptoms. Studies in musculoskeletal pain highlight the value of natural course designs for understanding typical trajectories of improvement, persistence or deterioration, particularly in the absence of targeted interventions.^{29 30}

Consistently, recent evidence emphasises that the literature on TMD has largely prioritised evaluating therapeutic approaches, such as interdisciplinary physiotherapy and dental interventions, rather than natural history, thereby perpetuating this evidence gap.³¹ Natural course studies should report descriptive outcomes, such as the proportion of patients who improve, worsen or remain stable, without attempting causal attribution.^{32 33} The absence of such data represents a significant evidence gap and limits the development of robust prognostic models.^{34 35}

The nine factors negatively associated with TMD outcomes were pain-related, especially pain during movement or provoked by palpation. In particular, joint pain with sound and muscle pain during movement showed stronger associations, broadly consistent with chronic musculoskeletal pain literature.³⁶,³⁸

In adjusted analyses, age, sex, depression and somatisation were not consistent prognostic factors. By contrast, pain-related clinical signs (eg, joint pain with sound; muscle pain on movement or palpation) appear more informative over the short to medium term. Heterogeneity in functional measures (JFLS, MFIQ, self-efficacy) and the lack of unadjusted estimates for function-limited quantitative synthesis likely contributed to effect variability.

Emerging evidence suggests condition-specific psychosocial profiles: anxiety/somatisation is more closely linked to TMD myalgia, and depression/persistent anxiety to migraines.³⁹ This may explain discrepancies with earlier studies^{7 40} and indicates context-dependent effects. Although female sex is associated with TMD prevalence, particularly in working-age women,^{5 41} the present review does not support a clear prognostic role for sex.

Important domains were under-explored: sleep (sleep quality, sleep bruxism), genetics, posture and cervical involvement. Sleep disturbances/bruxism show inconsistent associations due to diagnostic limitations. In particular, studies investigating the diagnostic accuracy of portable devices for assessing sleep bruxism in populations with comorbid sleep apnoea have highlighted significant challenges in measurement reliability,⁴² further complicating the integration of sleep-related variables into prognostic models.^{6 43 44} Only one study assessed it unadjusted.²³ Educational level and parafunction were mostly unadjusted and not meta-analysed.⁴⁴,⁴⁶ Genetic influences remain unaddressed (eg, interleukin (IL)-1, IL-6 and COMT Val158Met), including neck pain and motor dysfunction, which are also highly prevalent in individuals with TMD and have been linked to mandibular dysfunction in several studies, warranting a prospective evaluation.

Methodological limitations temper the certainty of these findings. The QUIPS tool identified attrition and confounding as the most compromised domains, with most studies rated as having moderate^{1314 26},²⁸ to moderately high risk of bias.^{23 24} Future research should apply more rigorous designs, including improved follow-up strategies, multivariate adjustment and handling of missing data through techniques such as multiple imputation.^{32 33}

The adapted GRADE assessment also highlighted imprecision (wide CIs that often cross the null) and limited replication across prognostic factors, thereby restricting the overall quality of the evidence. Nonetheless, most studies demonstrated low risk of bias in domains related to study participation, prognostic factor measurement, outcome measurement and statistical analysis and presentation, strengthening confidence in these aspects.

From a clinical perspective, the most consistent and actionable evidence suggests that pain provoked by movement or palpation may serve as a marker of poorer prognosis. At the same time, greater mandibular mobility and shorter symptom duration appear protective for functional recovery. Given the low certainty of the evidence, clinicians should emphasise individualised assessment based on symptom presentation and adopt shared decision-making when discussing prognosis and treatment expectations with patients, rather than relying solely on demographic factors.

Strengths and limitations of this study

A key strength of this review lies in its methodological rigour, encompassing a comprehensive search strategy, structured quality appraisal and transparent data synthesis. Throughout the process, measures were undertaken to minimise bias and enhance the robustness of the findings.

Although definitive conclusions regarding the prognosis and prognostic factors for pain and function in TMD cannot yet be drawn, this review identifies critical evidence gaps that future research should address to advance the field. It also underscores the paucity of impactful, high-quality studies in this area; by delineating these limitations, we aim to encourage more rigorous and clinically relevant prognostic research in TMD.

The main limitation of this systematic review is the small number of included studies and their overall methodological quality. Moreover, heterogeneity in follow-up durations and in the definitions and ranges of prognostic factors limited our ability to draw consistent conclusions about TMD prognosis.

Finally, we could not formally assess small-study or publication bias because fewer than 10 studies contributed to any single meta-analysis, in accordance with Cochrane guidance. Consequently, selective reporting cannot be excluded; this concern was reflected in the adapted GRADE certainty ratings for the relevant factors.

Overall, most studies were rated as having moderate to moderately high risk of bias, underscoring the need for more rigorous methodological strategies to reduce attrition and improve control for confounding factors in future prognostic research. However, these findings should be interpreted as associations rather than causal relationships, as the observational design of the included studies does not allow causal inference.

Clinical implications for treatment planning

Although the evidence is associative rather than causal, these prognostic indicators can support clinical reasoning for risk stratification and treatment planning. Patients presenting with higher pain intensity and pain provoked by movement or palpation may benefit from early, multimodal care (education, graded jaw exercises, analgesic optimisation and close follow-up). The presence of psychological factors (eg, anxiety, depression, somatisation, low self-efficacy) should prompt early screening and, when indicated, referral for multidisciplinary management, including psychologically informed physiotherapy and/or cognitive–behavioural interventions. Conversely, shorter pain duration and greater pain-free opening suggest a more favourable trajectory and may justify a stepped-care approach using conservative measures. These applications should guide shared decision-making rather than deterministic prediction, given the low-to-very-low certainty of much of the evidence.

Implications and future directions

This systematic review synthesised evidence on prognosis and prognostic factors for pain and functional outcomes in individuals with TMDs. Nine prospective studies were included, and 56 variables were examined after standardising definitions across studies.

In adjusted analyses for pain, 38 variables were assessed: 2 were identified as potentially protective (pain duration and pain-free mouth opening). In contrast, nine—primarily pain provoked by clinical examination or movement—were associated with a poorer clinical course. Most remaining variables showed non-significant or small effects.

Regarding functional outcomes, three factors (maximal mouth opening, shorter pain duration and pain-free opening) were potentially protective, whereas eight were associated with a higher risk. Only three variables were eligible for meta-analysis: pain intensity demonstrated a significant association with worse function, whereas self-efficacy and number of disability days were not statistically significant.

Emphasis on methodological rigour in future research

Future studies should adopt more rigorous designs, including improved follow-up/retention strategies, prespecified analysis plans, appropriate multivariable adjustment for key confounders, robust handling of missing data (eg, multiple imputation) and transparent reporting. Larger, harmonised cohorts using core outcome sets are needed to enable robust random-effects meta-analyses, sensitivity (including leave-one-out) and subgroup analyses (eg, by TMD phenotype and sex), thereby strengthening causal inference and reproducibility.

Conclusion

The available evidence from nine prospective studies suggests that certain prognostic factors—particularly those related to pain provoked by movement or palpation—are associated with worsening pain and functional impairment in individuals with TMD. In contrast, shorter pain duration and greater pain-free mouth opening may be protective of functional recovery.

Given the observational design and moderate-to-high risk of bias across studies, these associations should not be interpreted as causal. This distinction between association and causation remains essential to avoid overinterpretation.

Overall, the findings align with evidence from other chronic musculoskeletal conditions, emphasising the prognostic value of clinical pain signs over demographic or psychosocial characteristics. There is a clear need for more rigorous longitudinal research that systematically investigates biopsychosocial, genetic and parafunctional factors, with appropriate control for confounding and standardised assessment of pain and function. This review provides a framework for future high-quality studies to clarify prognostic trajectories and improve patient management in TMD.

Data availability statement

Data sharing is not applicable as no datasets were generated and/or analysed for this study.